The main point he makes is to use energy consumption to show us what changes.

Even by 1850, rising coal extraction in Europe and North America supplies no more than 7 percent of all fuel energy, nearly half of all useful kinetic energy comes from draft animals, about 40 percent from human muscles, and just 15 percent from the three inanimate prime movers: waterwheels, windmills, and the slowly spreading steam engines. The world of 1850 is much more akin to the world of 1700 or even of 1600 than that of the year 2000.

To be more precise: “Consequently, the 20th century saw a nearly 40-fold gain in useful energy; since 1800 the gain was about 3,500-fold.”

Think about that. The amount of energy we have available now is mind boggling. World wide, it is as if 60 people working non-stop for one human being, in countries like germany, I consume so much energy as if 240 people were working for me night and day. And while we can save energy, remember that “is not the case with the more than 5 billion people whose energy consumption is a fraction of those affluent levels, who need much more ammonia to raise their crop yields to feed their increasing populations, and much more steel and cement and plastics to build their essential infrastructures.”

The steep reduction in global undernutrition means that in 1950 the world was able to supply adequate food to about 890 million people, but by 2019 that had risen to just over 7 billion: a nearly eight-fold increase in absolute terms!

Growing crops needs sunlight, or always needed sunlight, but we added non-solar forms of energy two centuries ago, that allowed for higher yields.

Producing wheat now takes less than two hours of human labor per hectare (compared to 150 hours in 1801), and with yields of around 3.5 tons per hectare this translates to less than two seconds per kilogram of grain.

In the last two centures, producing a kg of wheat moved from taking 10 minutes of human labor to 1.5 seconds. There is a huge part about fish and growing things and where the energy comes from. I won’t go into that but simply point out that without the artificial production of ammonium, invented in 1903, we would not be anywhere near our production of food and creating ammonium takes huge amount of energy.

The global population rose from 1 billion in 1800 to 1.6 billion in 1900 and 6.1 billion in the year 2000, and hence the supply of useful energy rose (all values in gigajoules per capita) from 0.05 in 1800 to 2.7 in 1900 and to about 28 in the year 2000. […] An average inhabitant of the Earth nowadays has at their disposal nearly 700 times more useful energy than their ancestors had at the beginning of the 19th century.

Now for the big energy uses:

In 2019, the world consumed about 4.5 billion tons of cement, 1.8 billion tons of steel, 370 million tons of plastics, and 150 million tons of ammonia, and they are not readily replaceable by other materials—certainly not in the near future or on a global scale.

…

Iron ore smelting in blast furnaces requires coke made from coal (and also natural gas); energy for cement production comes mostly from coal dust, petroleum coke, and heavy fuel oil. The vast majority of simple molecules that are bonded in long chains or branches to make plastics are derived from crude oils and natural gases. And in the modern synthesis of ammonia, natural gas is both the source of hydrogen and processing energy.

…

As a result, global production of these four indispensable materials claims about 17 percent of the world’s primary energy supply, and 25 percent of all CO2 emissions originating in the combustion of fossil fuels—and currently there are no commercially available and readily deployable mass-scale alternatives to displace these established processes.

…

Without ammonia, 4 billion people would not be able to be fed. And I know that companies building roads and other things with concrete are actually recycling concrete at a ever greater rate.

Now imagine that the developing world wants to move the same way as china.

Replicating the post-1990 Chinese experience in those countries would amount to a 15-fold increase of steel output, a more than 10-fold boost for cement production, a more than doubling of ammonia synthesis, and a more than 30-fold increase of plastic syntheses.

So …

The data are clear: between 1989 and 2019 we increased global anthropogenic greenhouse gas emissions by about 65 percent. Even when we deconstruct this global mean, we see that affluent countries like the US, Canada, Japan, Australia, and those in the EU, whose per capita energy use was very high three decades ago, did reduce their emissions, but only by about 4 percent, while Indian emissions quadrupled and Chinese emissions rose 4.5 times. […] the Paris accord did not (could not) codify any specific reduction targets by the world’s largest emitters, and it would, even if all voluntary non-binding pledges were honored (something utterly improbable), result in a 50 percent increase of emissions by 2050.

This is just a fact of other countries wanting it nice too.

Consequently, even if those countries whose standard of living is today where China’s was in 1999 were to achieve only a tenth of China’s recent growth, they would experience a 10-fold increase of car ownership and a 40-fold increase in air conditioners.

Also …

The most radical way to cut energy costs and the environmental impact of nitrogen fertilizers is to reduce how much is used: that option is available to affluent countries with their excessive food supply and waste—but hundreds of millions of stunted children, mostly in Africa, need to drink more milk and eat more meat, and that protein can come only from substantially increasing the amount of nitrogen they use in cropping. Just to drive this conclusion home, annual applications of fertilizers average about 160 kilograms per hectare of agricultural land in the EU and less than 20 kilograms in Ethiopia, an order-of-magnitude difference illustrating the huge development gap that is so often ignored in appraisals of global needs.

He also touched on Covid.

Except the next time, the risk will be significantly higher because the combination of natural aging and prolongation of life will greatly increase the share of people over 65 years of age. The UN projects that share rising by about 70 percent by 2050, and in better-off countries one person in four will be older than that. How will we cope in 2050 with a pandemic that might be more infectious than COVID-19, when in some countries a third of the population is in the most vulnerable category?

To end …

Nobody in 1945 could have predicted a world with more than 5 billion additional people that is also better fed than at any time in history—even as it keeps wasting an indefensibly high share of all the food it grows. Nor did anybody foresee a world that relegated a number of infectious diseases (most notably polio everywhere, and tuberculosis in affluent nations) to historical footnotes, but that cannot keep economic inequality from widening even in the richest countries; a world that is at once much cleaner and much healthier yet also more polluted in new ways (from plastic in the ocean to heavy metals in soils) and, due to the ongoing biospheric degradation, also more precarious; or a world suffused in instant and essentially free information that comes at the price of massively disseminated misinformation, lies and reprehensible claims.

Bummer. Still very good, and really worth to think about.

It’s a fun book to read. Let’s get into a few quotes as usual.

The key to a good life is not giving a fuck about more; it’s giving a fuck about less, giving a fuck about only what is true and immediate and important.

But more importantly, and you sometimes feel that with the problems we talk about.

The Feedback Loop from Hell has become a borderline epidemic, making many of us overly stressed, overly neurotic, and overly self-loathing. Back in Grandpa’s day, he would feel like shit and think to himself, “Gee whiz, I sure do feel like a cow turd today. But hey, I guess that’s just life. Back to shoveling hay.”

Put differently.

The desire for more positive experience is itself a negative experience. And, paradoxically, the acceptance of one’s negative experience is itself a positive experience.

And we are living in a social media world, where somebody always seems to have it better. So positive is very inflated.

The author “once heard an artist say that when a person has no problems, the mind automatically finds a way to invent some.” And remember, you have a choice on what to give a fuck about. You are inherently dissatisfied, that is a part of like.

“Life is essentially an endless series of problems, Mark,” the panda told me. He sipped his drink and adjusted the little pink umbrella. “The solution to one problem is merely the creation of the next one.”

So, if that is true… “True happiness occurs only when you find the problems you enjoy having and enjoy solving.” Because “Emotions are simply biological signals designed to nudge you in the direction of beneficial change.”

Problems may be inevitable, but the meaning of each problem is not. We get to control what our problems mean based on how we choose to think about them, the standard by which we choose to measure them.

So you need to move the goal post to be happy.

Pleasure. Pleasure is great, but it’s a horrible value to prioritize your life around. Ask any drug addict how his pursuit of pleasure turned out.

Now for deeper psychology to think about.

Our brains are meaning machines. What we understand as “meaning” is generated by the associations our brain makes between two or more experiences. We press a button, then we see a light go on; we assume the button caused the light to go on. This, at its core, is the basis of meaning. Button, light; light, button.

So. Question #3: Would being wrong create a better or a worse problem than my current problem, for both myself and others?

While it feels like you vs the world, it is just you vs yourself. He has an interesting example about frame of reference.

“I asked myself a simple question: “Would I rather make decent money and work a job I hated, or play at Internet entrepreneur and be broke for a while?” The answer was immediate and clear for me: the latter. I then asked myself, “If I try this thing and fail in a few years and have to go get a job anyway, will I have really lost anything?” The answer was no. Instead of a broke and unemployed twenty-two-year-old with no experience, I’d be a broke and unemployed twenty-five-year-old with no experience. Who cares?”

Life is chances.

Life is about not knowing and then doing something anyway. All of life is like this. It never changes. Even when you’re happy. Even when you’re farting fairy dust. Even when you win the lottery and buy a small fleet of Jet Skis, you still won’t know what the hell you’re doing. Don’t ever forget that. And don’t ever be afraid of that.

Recently read something interesting. Picasso, did something between 50.000 and 135.000 pieces of art, depending of how you count, and only a few hundred, maybe a very few thousand, are famous. Maybe 2%.

The victim, if he really loved the saver, would say, “Look, this is my problem; you don’t have to fix it for me. Just support me while I fix it myself.” That would actually be a demonstration of love: taking responsibility for your own problems and not holding your partner responsible for them. If the saver really wanted to save the victim, the saver would say, “Look, you’re blaming others for your own problems; deal with this yourself.” And in a sick way, that would actually be a demonstration of love: helping someone solve their own problems.

Good book. Read.

"Disregard the Doomsayers: The misguided belief that “it’s too late” to act has been co-opted by fossil fuel interests and those advocating for them."

Now that is another good start, right from Page 6 and it’s good to remember that the “it’s too late” is now both used by climate people but actually mostly by climate deniers, which you probably, his words, do not even have to talk to as they are unconvincable.

“Changing the System Requires Systemic Change: The fossil fuel disinformation machine wants to make it about the car you choose to drive, the food you choose to eat, and the lifestyle you choose to live rather than about the larger system and incentives. We need policies that will incentivize the needed shift away from fossil fuel burning toward a clean, green global economy. So-called leaders who resist the call for action must be removed from office.”

This one is important to remember. Systemic change is important, but many of the disinformation people want to make it about your personal choice, because you might not want to, simply to avoid systemic change, so we have to or rather it makes economic sense.

“Doubt is our product, since it is the best means of competing with the “body of fact” that exists in the minds of the general public. —Unnamed tobacco executive, Brown and Williamson (1969)”

We are in a world of feelings, not facts and moving closer to it. This has actually been said many years ago.

“Nearly two decades ago, in his book The Demon-Haunted World, Sagan presaged with some trepidation the world we now live in: I have a foreboding of an America in my children’s or grandchildren’s time—when the United States is a service and information economy; when nearly all the manufacturing industries have slipped away to other countries; when awesome technological powers are in the hands of a very few, and no one representing the public interest can even grasp the issues; when the people have lost the ability to set their own agendas or knowledgeably question those in authority; when, clutching our crystals and nervously consulting our horoscopes, our critical faculties in decline, unable to distinguish between what feels good and what’s true, we slide, almost without noticing, back into superstition and darkness.”

Deflection campaigns are long known, e.g. when something like the bottle bill in the us was going to be introduced, doing something like Pfand in germany (you get money to bring back a bottle, which ads a lot of financial burden to the producer), special interest groups tried to imply that it is all about individual action, as when kids would just pick up bottles and cans. It’s to deflect from the systemic change.

Individual action doesn’t fund high speed railways, climate research or build renewable energy. Not at scale. It requires both individuals doing it, but supported by systemic change.

“When the climate discourse devolves into a shouting match over diet and travel choices, and becomes about personal purity, behavior-shaming, and virtue-signaling, we get a divided community unable to speak with a united voice. We lose. Fossil fuel interests win.”

If you ask the average joe on the street what they can do for climate change, it is not lobbying politicians, asking for change, but rather to stop eating meat, not fly, and turn off the lights.

Did you know, Greensburg in Kansas, was leveled by a tornado and has been rebuilt 100% on renewable energy … imagine knowing that. All the “it’s not possible” would be a lie.

The simple thing is this:

“Boycoff has a T-shirt (inspired by the work of climate communication expert Ed Maibach of George Mason University) that reads: “It’s real; it’s us; experts agree; it’s bad; there’s hope.”4 Note once again the carefully calibrated balance of urgency (“it’s bad”) and agency (“there’s hope”).”

From then on it gets complicated. There are a lot of good examples in the book, so I highly recommend reading it to get a balanced view.

Here’s what came across.

The agentic SDLC moment, arriving from all sides

The headline of the week — Andrej Karpathy joining Anthropic — was the kind of news that lands as a punctuation mark, not a surprise. Puja noted that Karpathy himself has said he’d like to spend time at a frontier lab as an advisor and may not stay long, but the optics still count: from OpenAI, to Tesla, to Anthropic, the centre of gravity for the people thinking hardest about LLMs has shifted again. Jonas‘s read was the funniest one — “he saw /goal and took it literally” — and possibly also the most accurate.

The substance that made the week feel coherent came from Puja’s piece on the software inflection point — agents, OSS, and innersource. The argument is that agentic SDLC isn’t only changing how code gets written; it’s changing what “open” and “shared” mean inside a company. If an agent can read every internal repo at once, the wall between “our code” and “open source we depend on” gets thinner, and the question shifts from “what should we open source?” to “what capability gates do we encode, so the agent stays on the right side of the line?” Worth reading slowly. I read it multiple times. It’s worth it.

Sitting alongside that: a repackaged clip from Karpathy framed around “the death of vibe coding.” Timo rightly flagged it as a clickbait edit of a longer Sequoia interview (the original is here), but the underlying point is real and worth sitting with. Martin‘s reply was the most generous treatment of it I’ve seen: he’s been building a full KDE AI studio outside work as a way to understand what agentic development really looks like, and his summary is that we’re already past tinkering. The job now is “architectural guidance and UAT” — telling the agent what didn’t pass, why the approach was wrong, what you missed in planning. And that, he points out, requires a remarkable amount of documentation, role definitions for different agents (researcher, planner, dispatcher, implementer, reviewer, fixer), and patience. “Agentic development is a real skill that takes a lot of time, effort, and patience to get right.” That’s the line I’m keeping.

And then in a separate thread — kicked off by Oreoluwa sharing applied-llms.org — Timo delivered the cleanest one-paragraph reframe of where we actually are: “RAG is irrelevant these days. It’s all about context engineering and the harnesses do a lot of the heavy lifting.” Your prompts no longer have to be precise — the harness asks you the questions. Jonas piled on with the practical reading list: muster workflows, /context management, the Karpathy loop, the new /goal feature. If you wanted a single short reading list to catch someone up on where serious AI-assisted engineering sits in mid-2026, that thread is it.

The token bill comes due

Three posts that look unrelated until you stack them.

One: the OpenClaw founder burned $1.3M in tokens in 30 days, and admits that without fast mode it would have been 70% less. Timo‘s reaction was the operator’s one — he has OpenClaw running in a VM on his cluster, and “it just wastes a lot of tokens.” Fernando added the line worth quoting: tokens have a cost — not only economic, but environmental. The bragging tweet shows a real disconnect.

Two: Amazon is setting per-employee targets for AI token use. Marian caught the right Goodhart quote (”when a measure becomes a target, it ceases to be a good measure”), and that’s exactly the failure mode. The moment you incentivise tokens, you stop optimising for outcomes and start optimising for the wrapper around them. We’ll see a lot of this in the next year and almost none of it will produce better software.

Three: the Google / Antigravity pricing thread, where Martin walked through the absurdity. Gemini had 24h usage limits; Antigravity has 5h limits, but those aren’t the same as the app’s 5h limits (until they are). There’s a new €100/month plan if you need to upgrade, and additional tokens you can buy that may or may not roll over. Puja added the detail that lands the point: AI Pro now bundles YouTube Premium Lite, but only for the account holder, not the family. We live in a shitty future, indeed.

Stack the three together and the picture is clear: the economics of agent use are still being invented in public, and the incumbent platforms are reaching for every lever — token caps, pricing tiers, internal targets, bundled side-perks — to find one that holds. None of them feel stable yet. The companies that figure out where in the stack to extract value without breaking their own product loop will be the ones that survive this phase.

Supply chain, again, and an honest response

The week’s most useful security read was Grafana Labs’ update on the TanStack npm supply-chain ransomware incident. Quentin‘s pointer was the right one: it’s notable not for the incident itself — npm supply-chain attacks are now boring in their frequency — but for how upfront Grafana was about what happened. The bar for “transparent incident comms” keeps moving, and the companies that meet it without flinching are doing more for the ecosystem than the ones that try to spin.

Adjacent: Cloudflare’s cyber frontier models post. Marian flagged the strategy tips for multi-agent orchestration buried inside it — worth a read if you’re building anything where multiple agents need to cooperate without trampling each other. And Henning’s first LinkedIn post on Shadow Agents hit the same nerve from the enterprise side: the problem of AI agents quietly being spun up inside organisations by individual teams, with no governance, is the next iteration of shadow IT, and it’s already happening.

What AI does to a workplace

Marian brought a Fast Company piece on what increased AI-driven speed costs in terms of organisational alignment and culture. His own observation was the most concrete part: every time you ask an agent instead of a colleague, that’s one less personal interaction — and over enough such substitutions, the connective tissue of a company gets thinner.

Puja‘s pushback was sharp and worth quoting because it’s the steel-man for the other side: most of what the author calls “non-alignedness” is really just asynchronous work and growth. AI transcriptions don’t make people skip meetings — they just replace half-assed note-taking with something slightly better. And speed of execution doesn’t cause misalignment; it forces you to fix the gatekeeping processes that were hiding it. Both reads have merit. The honest answer is probably that AI raises the stakes on culture work, in both directions — speed makes good cultures faster, and bad ones fail faster too.

What I took from this week

Two threads, one shape.

One: agentic development isn’t the future, it’s the present. The Karpathy news, Puja’s blog, the applied-llms thread, Martin’s KDE-studio side project — these are all the same story told four ways. The skill that matters now is not writing the code, and not writing the prompt either. It’s structuring the context: documents, roles, gates, harnesses. The companies that figure out how to encode their actual taste and constraints into that structure will get exponentially more out of every model upgrade. The ones that don’t will keep buying tokens and producing slop.

Two: the token economy is in a transitional, slightly ugly phase. OpenClaw burns $1.3M and brags about it; Amazon makes spending tokens a KPI; Google reinvents its pricing tier weekly. None of these are equilibrium states. The interesting question isn’t which provider wins — it’s which abstraction layer captures the value once the dust settles. Smart money says it’s the one closest to the human, not the one closest to the GPU.

If any of these sent you somewhere interesting, reply and tell me where you ended up. The follow-on conversations are the part I’m here for.

First off I need to start by that came out just now, around how to value ServiceNow. First of all, I can only suggest subscribing no matter what, but Alex has touched upon some good things in it, namely that we are moving away from just Chips and LLMs being the narrative, but rather, that we now need to move to real value being created on the user side. I just did a quick take on it on my LinkedIn.

It’s fits with something that we came across in Giant Swarm around Shannon’s piece on the seven things enterprises actually want from an MCP gateway. Obot is reportedly seeing a lot of meetings on this in the US — the market is warming up fast and the requirements are starting to crystallise. For those not knowing what MCP is, think of it as an API to a Service with Context for the AI to work with. It fits to the ServiceNow narrative and in short, has these specific points that companies are looking for and based on our experience, I can only agree:

• vendor-neutral control plane (multi-model, multi-provider): You will switch between LLMs or Providers, you need your MCP Model to be seperate

• identity-anchored access (SSO, auth), tool-level permissions, content filtering at the gateway (PII and worse): And this is not easy and we are banking on standards that will push this forward. You want your agents to be specific, and so you want the access to be as well.

• skills as first-class objects: They need to be as far model independent as possible.

• audit logs that reach the SIEM: ok, I think I don’t need to explain why audit logs are important.

• and deployment flexibility across hosted, private, and on-prem: Because you will have private data and want to use a local (open-source) LLM for it. Long term, you will want to be able to move your things freely between LLMs for performance and cost reasons.

I will add one thing to it that we feel is very powerful. Our MCP Product has a Workflow Engine that allow it to build code on things it learned, like new limited MCPs, that will allow it to query those with a lot less context, saving considerable LLM costs and creating more reliable outcomes. You need your agents to be free to do research, but some things are clear and can be moved to code that is a lot easier to run and a lot less variable.

A more fun one: the first AI boss launched in San Francisco. Andon is an autonomous AI agent operating as a manager, and it has already hired two real humans to work for it. Yes, really. Let’s see how this plays out.

The structural one: GitLab Act 2, the CEO's open letter explaining how the company is rebuilding itself to be ready for AI-driven software engineering. It's a piece of corporate self-positioning, but the transparency is unusual — laid out as a letter to customers and investors, with what they think their advantages are and what they're planning to build next. Zach's reaction was the honest one: hard to tell from outside whether GitLab is at war internally over any of this, but the willingness to be this clear in public is itself reassuring.

Security Incoming

The big one that I will have to think a bit about is the security changes that are afoot. With newer AI models that are in trials, the among of security problems we find it mind blowing. The reason is that these models are good at chaining different things together, instead of just exploiting one CVE.

As one example, we had a supply chain attach and yes NPM is a landmine, again — Aikido flagged another supply-chain incident in the npm ecosystem. Dominik's line was the one I'm keeping: "NPM always was a landmine. Now we just have a metal detector." That's roughly the current security posture for half the package ecosystems we depend on.

I’ll go into more detail on things in the future, but the general direction and velocity will lead to some good headlines in the near future.

Here’s what came across.

Agentic networking, and the politics behind the plumbing

The week’s most useful thread sat around kube-agentic-networking — a new Kubernetes SIG project for agent-to-agent and agent-to-MCP connectivity, with permission control on the routes. Adjacent to AI gateways like agentgateway, but pitched at a different layer.

Quentin framed it cleanly: this looks like the control plane (think Gateway API CRDs), with agentgateway as the data plane (think Envoy Gateway). Useful mental model. The tension underneath is that agentgateway is the best tool in town right now for our use case — but if the CRDs land and agentgateway doesn’t pick them up, you’re staring at a migration you didn’t ask for. Right now it’s only doing identity, so the surface is still small.

Then there’s the contributor mix. Quentin counted heads on the project: one from Solo, one Red Hat, one IBM, three unaffiliated, and ten from Google. The honest read is — better than a single-vendor project, still not as diverse as you’d want for something this load-bearing. Solo’s track record on adjacent projects is mixed: they move fast, which is good, and they sometimes make decisions other vendors don’t agree with, which is sometimes good and sometimes a migration tax later. Worth watching closely.

In the same neighbourhood: Project Capsule surfaced via a potential customer — a multi-tenancy automation layer that helps manage teams and namespaces. Not new, but the pattern of customers reaching for purpose-built multi-tenancy on top of stock Kubernetes is itself a signal.

And ing-switch — Sayam’s pre-KubeCon migration tool for moving off ingress-nginx — got a re-mention with a customer migration in mind. It already covered more annotations than the “upstream” tool back at KubeCon, and ships with a UI on top. Quietly useful for anyone still working on moving off nginx-ingress.

AI tooling — more headroom, fewer excuses on security

A few things in the AI/dev-tools lane, lining up in roughly the same direction.

First, Claude usage limits got doubled. Real-world report from inside the Swarm: ran a customer update plus a new version of a skill in one session, and didn’t come anywhere near where the old limits would have bitten. That’s a meaningful working-conditions change for anyone running AI-heavy workflows on a paid plan.

Second, Vercel introduced DeepSec — a security harness that sits over Claude and Codex to find and fix vulnerabilities in the codebase. Early word in the channel was positive enough to put it on the “try it” list. The category — agent-aware static analysis with a fix loop attached — is starting to look real rather than aspirational.

Third, Anthropic’s workload identity federation now supports SPIFFE. Quietly important. If you’ve been waiting for a clean way to give workloads (not users) scoped Claude access without baking long-lived credentials into pipelines, that’s the missing piece.

And on the meta-question of how anyone is supposed to use all this: Kaspar von Gruenberg’s piece on the four levels of agentic software development. Useful framing. Most of us, honestly, are still living somewhere between level 0 and level 1 — and the leap to level 3 is less about the tools and more about being willing to actually delegate work that used to feel non-delegatable. Not just for code, either; the same staircase applies to most knowledge work. At the same time is the reason we are building both the access as well as the agent platform, based on Open Source and developing standards. It means the agents need to move into K8s Servers and have a Synchronization Layer with Governed Access to do their real work.

Fun stuff from the deeper end

Kubeshark v2 — the Wireshark-for-Kubernetes tool is shipping a v2 and the preview looks promising. Worth a look if you’re tired of squinting at tcpdump for service-mesh traffic.

What I took from this week

Two things.

One: the agentic-networking layer is being designed in the open, mostly by one company, and the rest of us are along for the ride. That’s not unusual for emerging Kubernetes territory — but it’s worth being honest that “vendor diversity in the contributor list” is an actual product feature, not a nice-to-have. The Quentin/agentgateway tension is the early version of a conversation we’re going to be having for years.

Two: when AI tooling gets quietly better — doubled limits, SPIFFE-based identity for workloads, security harnesses with real fix loops — it doesn’t feel like news, but it changes what’s possible to do on a Tuesday afternoon. The boring infrastructure improvements are the ones that compound.

Here’s what came across.

Warp goes open source — and reinvents what a contribution is

The biggest thread of the week came off Warp’s announcement: the AI-native terminal is going open source. The interesting bit isn’t the licence — it’s the contribution model. Their pitch is that “Oz” (their AI agent) plans, implements, and reviews features, so a contribution can effectively be a well-formed prompt.

Puja liked this immediately, and not just for terminal reasons: he raised the broader question of whether this becomes a model for inter-team contribution at companies like ours. What if one team had its AI pipeline set up so another team needing something — say, a cluster release tweak — could just file an issue and have it implemented? Push that further and account engineers, or even customers, could “contribute” feature requests directly into the codebase. Long-term thinking. Worth chewing on.

Martin pushed back where it deserves pushing: a single model planning, implementing, and reviewing its own work fails in practice. He’s tried it with Gemini and Codex on real features — they don’t catch the flaws in their own designs. His current setup splits the work: Claude for planning and review, Gemini/Codex for implementation. That separation is doing real work. Same-model self-review is one of those ideas that sounds clean and falls apart on contact with subtle bugs.

Puja’s other note was about Warp interfering with Claude Code in subtle ways — minor glitches, plus the friction of Warp wanting to use its own AI backend rather than the one the team has paid for and approved for on-call use. That’s a sleeper category of problem with AI-bundled tools: the model lock-in shows up as UX friction long before it shows up as a billing line.

In general I really like the idea as something to think about how we might want to get external contributions to our open source repositories. OpenClaw actually does it that way in some form in that the founders wants to get your prompts AND your code. I would love to hear your feedback.

Surveillance, tokenmaxxing, and the slow trust drain

Two links and one continuation, all pulling in the same direction.

First, Meta is starting to capture employee mouse movements and keystrokes as AI training data — yes, really. Zach‘s flat observation cuts to it: “newsworthy because it’s Meta, but many smaller companies will do the same.” That’s the bit that keeps me up. The headline is one company; the trend is the whole employer–employee data relationship being quietly rewritten under the cover of “AI training.”

The follow-up came a day later: Pragmatic Engineer’s piece on “tokenmaxxing” — engineers burning tokens to game internal AI-usage leaderboards. Jonas was on a tear in this thread (eleven replies, mostly his), and the throughline was sharp: tokens used is a stupid KPI. It’s the new “lines of code” or “calories burned.” It enhances the wrong thing. The deeper problem is that you can’t measure AI adoption with raw inputs; you have to measure outcomes, and outcome measurement is genuinely hard, and so companies grab the easy proxy and reward the wrong behaviour. He’s right. This is policy work, not tooling work.

Third in the cluster: Socket flagged a Bitwarden CLI supply-chain issue. Puja took a small victory lap for having moved off Bitwarden a while back. Simon is still on it but doesn’t use the CLI; Martin rolled his own read-only CLI after using the official one exactly once and hating it. The pattern across all three replies: people don’t trust the official tooling enough to use it as shipped. They wrap, replace, or avoid. That’s a quiet signal.

String these three together and the picture is consistent — the ambient level of trust in the tools that touch our keystrokes, our credentials, and our productivity metrics is degrading, and most of us are coping by writing our own thin layer.

Where infrastructure goes when nobody’s looking

A small cluster of posts about places where the substrate is moving — not loudly, just deliberately.

Mitchell Hashimoto wrote about why Ghostty is leaving GitHub. Quite a statement from the person behind half the HashiCorp tools you’ve ever used. The TL;DR (per Puja): he loves GitHub, but the outages have made it hard to work on, and he doesn’t believe they’ll get their act together. When someone of Mitchell’s stature with that kind of operational eye writes the move-out post, others will read it and think.

On the other end of the scale, the Dutch central bank picked Lidl for its European cloud. The reactions in the channel were a clean 4-emoji “uh-oh” / 2-emoji “eyes.” Whatever you make of the choice itself, the fact that European sovereign-cloud procurement is now this wide a field — including a discount supermarket chain’s IT arm — tells you something real about the appetite to escape the hyperscaler default.

And on the policy frontier, the UAE announced 50% of government services should run via agentic systems — though as Robin caught, the original target is two years, not five. Jonas‘s take: of the places willing to actually try this, the UAE isn’t the worst lab. I agree. We’re going to learn something useful from how that plays out, regardless of whether they hit the number.

Fun stuff from the deeper end

The week’s grab-bag:

hl — a CLI log parser that handles JSON and logfmt and falls back to passthrough for unknown formats, so it never silently swallows a line. Nice property. Posted with a screenshot.

getkloak.io — a clever use of eBPF, dropped by Simon. Jonas‘s honest assessment: “this software adds a malware-like level of complexity to networking which I imagine could be very confusing and hard to debug. Nevertheless, very impressive.” That’s the eBPF trade-off in two sentences.

cpg — generates CiliumNetworkPolicies from denies in the Hubble relay. The kind of small-but-pointed tool that the Cilium ecosystem keeps producing. If you’re running Cilium, take a look.

google/skills — Google’s take on the skills concept for admin tooling. Jonas has been kicking the tyres and reports mixed results: random errors, misunderstood commands, auth issues. His broader point — Google’s API surface is so large that a curated skills layer is genuinely useful — is the right framing. The execution is still catching up.

NVIDIA’s free hosted coding models — for private projects, hobby work, or when you’ve burned your subscription limits. Pairs nicely with opencode. Standard caveat from Puja: with free models, never feed in anything you wouldn’t post publicly.

Claude’s tokenizer is more expensive on non-English — dropped without commentary, but the implication is loud. If your product runs on Claude and serves a non-English market, your unit economics aren’t what you think they are. Language is not meaningless, even with AI.

A six-fingered AI-detection trick — a criminal-AI-detection meme: add a sixth finger to your photo and AI tools flag it as AI-generated. Funny in a depressing-arms-race way.

King Charles, briefly — geopolitics is now world leaders trading lines that used to be reserved for 3am bar arguments, as Luca put it. Hard to disagree.

What I took from this week

Two things, mostly.

One: the AI-tooling conversation has moved past “is it useful” into “what is it doing to our incentives.” The Warp thread, the tokenmaxxing thread, the Meta surveillance thread — they’re all the same conversation in different costumes. We’re rewiring how people contribute, how their work is measured, and what data is harvested in the process. The companies that get this rewiring right will quietly outperform the ones that just chase the leaderboard number.

Two: there’s a real, slow exodus underway from the defaults — GitHub for code, US hyperscalers for cloud, official password-manager CLIs for credentials. Each individual move is small. The aggregate is a story.

And three, just a small one: the channel is at its best when someone disagrees in public. Martin pushing back on single-model self-review, Jonas arguing through the tokenmaxxing implications eleven replies deep — that’s where the work happens. Posting a link is cheap. The disagreement is the value.

If any of these sent you somewhere interesting, reply and tell me where you ended up. The follow-on conversations are the part I’m here for.

We have a channel at Giant Swarm called #swarmalicious. It’s exactly what it sounds like: a link dump where people drop the things they stumbled across, with a sentence of context if they’re being generous. It’s been running since 2014, and honestly, it’s still one of the best signals I have for what the team is thinking about.

Here’s what came across the channel this past week — and the small debates that spun off around each link.

The Cursor valuation thing

I dropped this X thread about Cursor’s latest valuation with about as much commentary as I could muster: “this world is crazy.” The thread that followed was better than the post.

Pau kicked it off with “wild, never seen something like this tbh.” Puja‘s take: with their own models finally getting good enough at coding, Cursor needs a harness they can specialize to their own model — and that’s what the valuation is buying. Robin pushed back: the IDE is a VS Code reskin, and Composer 2 is based on Kimi K2.5 — how does that get you near $10 billion? Fernando reminded everyone that valuation is mostly about traction and user count, not quality (he brought up Wallapop at €600M burning cash and making nothing — a useful gut-check). Puja closed it out: “yeah, Cursor is still one of the leaders in the space.”

I don’t have a clean answer. But the question “what exactly are you paying for in an AI IDE?” is one I think about often, and the honest answer keeps shifting.

Anthropic and the quiet browser extension

A Golem article came up (with the original English write-up on thatprivacyguy.com) claiming Claude Desktop installs a browser extension native-messaging host without asking. Marian checked his own machine — yep, com.anthropic.claude_browser_extension.json sitting in every browser’s NativeMessagingHosts folder. Chrome, Chromium, Opera, Edge, Arc, Vivaldi, Brave. Silently.

Robin pointed out this isn’t unheard of — 1Password does the same thing, dropping config files into browsers that aren’t even installed. That’s a fair counter. But Marian’s reply stuck: “I agree with the author that something like this shouldn’t be done quietly. It should only be done with user consent.”

That’s the line, I think. The behavior isn’t necessarily malicious. But “we’ll quietly write to your filesystem because it’ll make the UX nicer” is exactly the kind of default that erodes trust over time. Especially for a company that’s positioned itself around safety.

Gemini gets a native Mac app (apparently vibe-coded)

The link to Gemini’s new native Mac app got dropped. Marian assumed it was another Electron wrapper — “not sure an Electron app is any better than a browser tab” — then did a double-take when Puja pointed out it’s native Swift. Marian: “Oh-kay, that is so unexpected these days, I didn’t even notice.”

Then the interesting bit: Google said this was vibe-coded with their Antigravity tool. Puja ran with it — what if agents kept their primitives (the things they understand well) and had skills that compiled those primitives into native frameworks like Swift? Same code, compiled differently depending on the target. He pointed to projects that use Docker but ship a skill that converts to Apple containers to run natively on Macs. Marian raised the obvious wish: a cross-platform framework for native GUI apps that actually works.

That’s a shape I’d bet on. The world doesn’t need another Electron app. It needs a way for agents to target native without the human having to rewrite everything.

Worth noting alongside this: Anthropic shipped Claude Design for Mac this week too, off the back of the Opus 4.7 upgrade. The Mac-as-first-class-target push is happening on multiple fronts.

IPv8 (please no)

“Forget IPv6, IPv8 is the new shit.” The reactions were mostly sighs. Jonas was blunt — we’re twenty years too late for a competing protocol with an easier migration path, not when IPv6 rollout is finally progressing. He did float an alternative: what if nations or blocks (EU, China) just pushed official services to IPv6-only? Force the migration by cutting off the old path.

Simon noted that IPv8 is literally just a proposal someone submitted. Anyone can submit one. It doesn’t mean anything yet.

So: not a story yet. But a good reminder that protocol politics is a slower-moving beast than most people assume.

Fun stuff from the deeper end

A few links that aren’t about AI for once:

marge — a new tool for approving and merging PRs, shared with the line “sometimes you have to like a tool just for the logo.” Fair. The dolphin is good. Jonas wondered how it compares to our internal devctl pr approve-merge-renovate; Simon and Jose agreed it’s essentially a more generic version of the same idea (and better than Simon’s “crappy bash function using the GH API directly”).

smolvm — worth experimenting with. Jose mentioned several friends are using Lima and loving it; Simon looked at Lima but prefers smolvm’s party trick — you can package a VM into a standalone file and share it. That’s a nice property.

RAM has a refresh flaw — a deeply nerdy but fascinating video: DRAM periodically stalls during refresh, blocking reads. Most of the time you’d never notice, but in ultra-low-latency systems it shows up as tail latency spikes measured in nanoseconds. Jonas: “Not all RAM is created equal. (I ♥ SRAM.)”

The ASIC unbundling — dropped without commentary, which in the channel usually means it’s worth reading.

drawio-mcp and cve-mcp-server — two more MCP servers for the pile. The first lets agents draw diagrams. The second surfaces CVEs. Neither life-changing on their own, but the MCP ecosystem is quietly getting dense.

Opus 4.7 best practices — worth a skim if you’re using Claude Code daily.

zerodayclock.com — a countdown clock for zero-day patches. Mostly vibes, but I like the vibe.

Itay is leaving Aqua — one of the sharper people in cloud native security, moving on. Worth watching where he lands.

What I took from this week

Three things.

One: the AI tooling category is in an awkward middle period. Big valuations, lots of harness engineering, and the question of where actual durable value lives is still open. Not scary. Just unsettled.

Two: the quiet-install pattern — whether it’s Anthropic or anyone else — is going to keep being a trust issue, and the companies that get it right will be the ones that ask first. Every time.

Three: the best link dumps still come from people who know what they’re looking at. No algorithm I’ve tried beats a dozen engineers with taste and a shared channel.

If you’re in a team that doesn’t have a #swarmalicious equivalent — start one. You’ll be surprised how much it compounds.

If any of these sent you down a rabbit hole, reply and let me know where you ended up. That’s the part I find most useful.

I know that sounds strange coming from someone who just launched an AI product. But bear with me.

We’re seeing a pattern with enterprises adopting AI for operational work. It tends to follow three stages — and most companies get stuck at stage two.

Stage one is connecting. You give AI agents access to your data and systems — governed, scoped, audited. This is the “can we even do this safely?” phase. It’s necessary, but it’s not where the value lives.

Stage two is automating. AI agents start doing real work: analysing logs, responding to events, running workflows. This is where the excitement is right now. You deploy agents, they handle tasks that used to require a person, and suddenly your team has more capacity. Good.

But here’s the thing: if you stay at stage two, your AI costs keep growing. More agents, more tokens, more API calls. The work gets done faster, but the cost curve doesn’t bend. For many enterprises, this is where the disillusionment sets in — “I thought AI was supposed to reduce costs?”

Stage three is where it gets interesting. This is what we call evolve.

An AI agent that does the same task a thousand times should eventually notice the pattern. It should be able to say: “This is what I do every time. Here’s the rule. Here’s the workflow. You don’t need me anymore for this.”

At that point, you take the agent out of the loop. No more tokens. No more latency. Just a deterministic process that runs — cheaper, faster, and more reliably than the AI version.

The best AI agent, in other words, is the one that makes itself redundant.

This isn’t a theoretical idea. It’s how we’ve designed our AI OS. Muster provides the governed access. Klaus runs the agents. But the system is built with the expectation that agents should be working towards their own retirement — identifying patterns, codifying them, and stepping aside.

Think about it like this: AI is the exploration phase. You use it to figure out what the process should be. Once you know, you don’t need AI anymore for that particular task. You need automation — simple, deterministic, cheap.

The companies that will see real OPEX reduction from AI aren’t the ones running the most agents. They’re the ones systematically converting agent work into automated workflows. It’s the difference between paying for intelligence every time and paying once to learn the answer.

This won’t happen overnight. Not every task can be codified. Some work genuinely requires the flexibility that AI provides. But for the large category of operational work that is repetitive, pattern-based, and rule-driven — the path is clear.

Use AI to discover the pattern. Then let the pattern run itself.

If you’re thinking about how to make AI reduce your costs instead of just adding a new line item — this is the conversation I’d love to have.

And the room goes quiet.

Here’s the thing most people get wrong about AI agents and data sovereignty: the question isn’t whether the LLM runs locally. The question is where the agent operates — where your data is processed, where decisions are made, where actions are taken, and who controls all of that.

An AI agent can talk to an external model. That’s fine. What matters is that the agent itself — the thing that reads your logs, touches your infrastructure, accesses your customer data — runs inside your environment, under your governance, with your permissions.

Think about it like a consultant. You might hire someone with expertise that lives outside your company. But you don’t send them all your data and let them work from their own office with no oversight. They come to you. They work in your systems. They follow your rules. And you know exactly what they did when they leave.

That’s the model we built Klaus around.

Klaus deploys AI agents into your Kubernetes clusters. The agents run continuously — analysing, responding, managing — on your infrastructure. They can use external LLMs (Claude, GPT, open-source models, whatever fits your requirements), but the agent itself lives where your data lives. Your data doesn’t get shipped somewhere else for processing. The actions happen locally. The audit trail is yours.

If you’ve used tools like Claude Code, Klaus is what that concept looks like when it’s deployed into enterprise infrastructure — running continuously, governed, and under your operational control.

The name is German — because we are. (It’s a hard K, by the way.)

This distinction — between where the model runs and where the agent operates — matters more than most vendors want to admit. A lot of AI tooling today bundles everything into a hosted service: the model, the orchestration, the data access, the actions. That’s convenient. It’s also a non-starter for enterprises with compliance requirements, data residency obligations, or simply a security posture that doesn’t allow sensitive data to leave their perimeter.

With Klaus, you choose the model. You control the data. The agent runs in your cluster. And if you ever want to run a local LLM as well — because your compliance requirements demand it, or because you want to reduce external dependencies — that path is open too.

Here’s what Klaus is not: a model provider. We don’t build foundation models. We don’t compete with OpenAI or Anthropic on that level. What we build is the operational layer — the part that makes it possible to run AI agents reliably, continuously, and under your control.

There’s a gap in the market that’s hard to see if you’re focused on model capabilities: the gap between “it worked in the demo” and “it runs reliably at 3am without anyone watching.” That gap is operations. And it’s exactly the gap Giant Swarm has always filled — first for Kubernetes, now for AI workloads.

Klaus doesn’t make AI magical. It makes AI operational. And in enterprise environments, that’s the harder problem.

If you’re exploring how to run AI agents on your own infrastructure — with the governance and reliability your organisation requires — I’d be happy to walk you through what we’ve built.

It’s “how do we let this thing touch our systems without losing control?”

I’ve had this conversation dozens of times over the past year — with CTOs, with heads of engineering, with platform teams trying to figure out how AI fits into their world. The excitement about what AI agents can do is real. But so is the anxiety about what happens when an agent has access to production data, customer records, or infrastructure controls.

And honestly? That anxiety is well-placed.

Most AI tooling today assumes a single developer on a laptop. You connect your agent to an API, give it some tools, and off it goes. That works great for a demo. It works great for a side project. It does not work when you have compliance requirements, multiple teams, sensitive data, and an auditor who wants to know exactly which AI accessed what, when, and why.

This is the gap we kept seeing. Not a lack of AI capabilities — a lack of governed access to enterprise systems.

That’s why we built Muster.

Muster is an enterprise MCP Gateway — think of it as the secure data plane for AI agents. Every request logged. Every permission scoped. Every connection controlled. It sits between your AI workloads and your enterprise data, making sure that agents can only see and do what they’re explicitly allowed to.

The name is German — “mustern” means to inspect, to review. It’s what you do before you let something through.

What Muster is not: a way to slow AI down. The goal isn’t to add friction for the sake of control theatre. It’s to make AI agents usable in environments where “just trust it” isn’t an option — which, in my experience, is most enterprises.

RBAC, audit logging, scoped permissions — these aren’t exciting features. They’re table stakes for any technology that touches production systems. The fact that most AI tooling doesn’t have them yet tells you something about where the industry is focused. (Hint: it’s not on the people who have to run things in production.)

We think governed access is the foundation everything else gets built on. Without it, AI agents in enterprise environments remain an experiment. With it, they become infrastructure you can actually rely on.

If you’re trying to figure out how to give AI agents access to your systems without giving away the keys — that’s exactly the problem Muster was built to solve. Happy to talk about what that looks like in practice.

Fair question. Let me explain why this isn’t the pivot it might look like.

For over a decade, we’ve been doing one thing well: running production infrastructure for enterprises that can’t afford it to break. Kubernetes management, platform engineering, operational reliability — the unsexy but critical work that sits between “the developer pushed code” and “the service is running in production at 3am.”

What we’ve learned from that work is that the hardest problem in enterprise technology isn’t building something clever. It’s making something clever run reliably, under governance, at scale, in environments where failure has consequences.

That problem hasn’t changed. The workloads have.

AI workloads are the new thing that enterprises need to run seriously. Not as demos, not as experiments — as production systems that touch real data, make real decisions, and need real oversight.

And the gap we see is familiar: most AI tooling is built by companies working their way down toward infrastructure. They start with models, add some orchestration, and eventually realise they need to figure out deployment, security, compliance, and operations. That “figuring it out” phase is where enterprises get hurt.

We’re coming from the other direction. We already know how to run things in production. We already have the trust of enterprises that need this done right. What we’re adding is the AI-specific layer — the governed access (Muster), the agent runtime (Klaus), and the evolution path from AI to automated workflows.

This is why we now have two complementary businesses:

giantswarm.io — our platform business. The Kubernetes management platform that enterprises and cloud providers rely on. Battle-tested. Production-grade. This isn’t going anywhere.

giantswarm.ai — our AI OS business. Muster, Klaus, and what comes next. Built on the same operational rigour as the platform, but designed for the age of AI workloads.

The foundation is the same. The operational expertise is the same. The customer relationships are the same. What’s new is the workload type — and the recognition that enterprises need someone they trust to run AI the way we’ve always run infrastructure: seriously.

I won’t pretend this isn’t ambitious. Building an AI Operating System on top of a platform business requires focus and execution. But the alternative — watching AI tooling vendors struggle to learn what “production-ready” means while our customers wait — didn’t feel like the right choice.

We know what keeps a CTO up at night. We know what “it works reliably at 3am” actually requires. And we know that the gap between a demo and production is exactly the gap we’ve always filled.

The job hasn’t changed. The workloads have.

If you’re wondering what a serious enterprise AI runtime looks like — one built by people who’ve been operating production infrastructure for years — I’d enjoy that conversation.

_The power of waves of the pacific is mind boggling._

_And Zapallar is just beautiful._

_Listening to the pacific at sun down is a must._

_I loved the nature in general and interplay of flowers and rocks and water._

_Atacama Desert is diverse beyond belief._

_Visited a wonderful Canyon in the Desert._

_Geysirs were thrilling._

_It looks like a different world._

_Welcome to Mars._

_Having breakfast in nature._

_The Pelikan’s at the Chiringuito in Zapallar_

I might post some more photos some time. For now, I think it gives you an idea of just a tiny amount of the wonders of the country.

We learned a few things everyone already knew

Before Giant Swarm & Friends, there was our Giant Summit in 2019. Picture it: a big giant hall in a BBQ shop filled with all our awesome customers for a party slash get-together. As you can probably guess, we did some barbecuing, and maybe less predictably, our customers spoke openly about their problems and amongst themselves found camaraderie and relief. While we offered the space to discuss, they drove the conversation. And so we discovered the first thing everyone already knew: a problem shared is a problem halved. Simply having the forum to discuss concerns seemed to provide intrinsic value. Whether a solution was discovered or not, a sense of family and a shared vision emerged, both of which lasted long after the Giant Summit had ended. And so, once again, we learned something that everyone already knew: casual conversations at parties are the glue that holds us together, and the world’s problems are often solved over a beer with friends. This isn’t because beer is magical* but rather because beer with friends simply cannot be faked. As my friend Manuel said as he delivered the closing speech at Pirate Summit: “You only connect with authenticity. Nobody wants to connect with fake.”

We struggled with everyone

Yup, COVID. No surprises there, but perhaps what was surprising at the time was how resilient our bond with our customers was (and thankfully still is). It mattered that we couldn’t all meet IRL, I’m the guy who tweets #realhugsrock, after all. But it didn’t matter enough to stop us from trying and succeeding at a bunch of remote events. From a remote Christmas party that included a magician and cocktail making to what is now our Giant Swarm & Friends events, we didn’t want to stop connecting with our community just because the world had stopped.

We learned from the struggle and thrived with everyone

Our Giant Swarm & Friends events were an attempt to rekindle the soft and fuzzy feeling of our IRL events while providing real value in terms of conversations and topics. Like our Giant Summit, we again wanted to center on our customers and their challenges. We invited our customers to talk about topics that were important to them rather than try to dictate what should be important to them and use it as a marketing stunt. As a result of this, we also decided to come up with some ground rules that support our mission.

1. Most importantly: Giant Swarm & Friends is not a sales event.

2. If we want to create a community, we don’t talk; we facilitate conversations.

3. Nobody’s special, so everyone’s special — first-name basis only.

4. Chatham House Rule

5. Everyone is welcome and encouraged to share.

6. Not exactly a hard rule, but chocolate, craft beer (alcoholic or nonalcoholic), and some stickers were shipped to all guests.

Regarding the last rule, we’ve been lucky to have our friend Daniel as our unofficial beer expert, who has guided us through the different types of beers in our selection. We’ve also been lucky that the conversations during the events have mimicked conversations IRL — in other words, they have been non-linear, organic, and often very entertaining. Our first event started at 8pm and finished at 1.30am, which is a point of pride, although we have since attempted to end them earlier as it’s still a weeknight. The debut included only about two hours of planned discussions, including presentations, Q&A, and beer talks. I can’t say I exactly remember the details of the remaining hours, but I do remember laughing, and I think that qualifies as a job well done.

We looked to the future

One of the best endorsements of our Giant Swarm & Friends events is that we have an almost 100% re-attendance rate. Those who join once are likely to join again, and for me, that’s a huge vote of confidence. We intend to reward that loyalty by continuously improving and building the community beyond the events. How that will all take shape is still being defined, but I do know it will all start with a conversation with our friends.

_*Although I know we have all felt otherwise at some point._

So it happened again this week, and as we have grown to close to 80 people, and 65 showed up, we needed a new system. The old system sadly does not work anymore. We used to rent one or several large houses and cook our own food, but while being a great social glue, it is too hard when you get to more than 30 people (but there are some ideas ;)). The last few events we have taken to renting a small boutique hotel, an island in croatia, and now, the Mas Salagros EcoResort.

_The view from the restaurant towards the main pool of Mas Salagros._

It was a great choice of venue, with wonderful rooms, great staff and food, and settled into a mountain valley. It had an amazingly tranquil vibe to it that rubbed off on everyone. But I thought I’d quickly write down what we do at these events and if you have questions, reach out.

People come mostly on monday, by train, plane or automobile, depending on where you live and what your preference is. Taking that monday as a day or arriving is actually good, because being a fully remote company, when suddenly everyone sees loved collegues in flesh, they want to talk and need time. So there is much rejoicing that first day and it is soooooo lovely to see everyone. On the first day I held a presentation before dinner showing lots of old photos and telling a bit about the history of Giant Swarm, which is fun for everyone, but insightful for all the new faces. The night continued with chats and dancing until well after midnight.

Tuesday is where the work starts… and at 10:00, right after breakfast, we met in our biggest meeting room, and started our Open Space. This time we had a big wall of subjects and a great mingling of groups, which is really what a well run Open Space helps to achieve: people that otherwise don’t talk to each other suddenly do, as the subject is interesting for both. We want people in the company to meet and form bonds and an Open Space is the perfect setting for that.

We had sessions on CAPI and how to work with Upstream projects of the CNCF. But also on how to make decisions, how to improve support, how to become a platform, leveraging our community approach, and many many more. I think the first day alone we had close to 20 sessions alone.

For the fun and games, we then, before dinner, had a runway show of all the old swag and to launch our newest items. Boy do we have some good models … I am wondering if the golden shorts need to become a swag item in general.

Wednesday is another day of Open Space and after dinner, at 21:00, we had a game of fun facts. Everyone needed to submit fun facts about them selves, one person gathered those and put them on slides, always giving an option of 4 people where that facts might fit. I now know who split their head open as a kid, recorded an album, got 3rd place a pole-dance competition, and many more. It’s a perfect setting to getting to know your team mates again.

Back on Thursday, we close the Open Space and everyone went back into their original teams that they normally work in to gather all the feedback and information and continue along their way. As a side note, the entire time, we are taking care of our customers, answering support requests, working on features, … we just do talk to our customers and ask if something can take until next week, and as we are just transparent and honest as we always are, this really works.

Thursday afternoon we had some events people could choose, like a wine tasting, nature walk, golf, … and then we had a nice last evening and with conversations and hugs and on friday, starting very early, the first people traveled back. A few are staying a bit longer as a holiday, but most are probably back home with their families by now.

All in all, it was a great success and I am feeling humbled to work with many so great people and am looking forward to many more onsites.

_Generated using Midjourney through the post title._

We at Giant Swarm have actively and conscientiously tried to rebel against this by contributing to our community through peer-to-peer engagements and events such as Giant Swarm & Friends. The name of the event we hope serves as a signpost of our intentions and a reminder to ourselves: a gathering of friends. As simple as that.

However, it’s not simple. We have to constantly be on the lookout for the traps of ‘good’ marketing. We don’t want to be good at marketing, or more specifically broadly targeted marketing speak, we don’t have to be. We want to be good at sharing what we do and connecting with our community in service of their interests, not ours.

The truth is that my intentions will always clash against reality — to maintain any connection that is real requires work, but how do I scale this when I aim to develop and maintain not just a handful of connections but upwards of 200? At the end of the day, Giant Swarm is a conduit for connection but not the connection itself.

People want to talk to people, and people build relationships with people and not with companies. This is not done in a one-dimensional way around one subject, e.g., Kubernetes. Instead, we are all multi-faceted, and we as a company need to enable conversations amongst individuals inside and outside Giant Swarm with the right facets. This means we must experiment, expand, and be creative, but it also invariably means we must make mistakes. After all, what is a more human and real experience than making a mistake? What do mistakes look like? Maybe, I should point out what they don’t look like. People vehemently disagreeing with us is a feature, not a bug, because it means we’re trying something. I would rather that than inspire no feelings at all. Inside Giant Swarm, we encourage honest conversations that sometimes get tough.

I’m inspired by our mission: to help companies innovate. We believe Kubernetes is currently one of the best ways to enable innovation, and so we have made ourselves Kubernetes experts to serve our community. We want to build a community because we believe that’s the best way to be generous with what we’ve learned and greedy for what we still need to learn. The bigger the community, the more everyone benefits. This is the same ​​reasoning that led us to ensure we have one product and not individual solutions on a per-customer basis, as everyone gains by fixing a problem experienced in the real-world working with one customer.

There is a lot more to say, but this is why this is a series, and I will be continuing on a biweekly basis. I would love to hear your thoughts on anything I mentioned here — but maybe don’t message me on LinkedIn; I might not be there much longer.

On Fertility: _The replacement level of fertility is that which maintains a population at a stable level. It is about 2.1, with the additional fraction needed to make up for girls who will not survive into fertile age._

which leads to one of the big ones. It seem that for the _during the 50 years between 2020 and 2070_ almost 75% of all births will be in Africa. Now this is something to think about on so many levels, and for me a founder and entrepreneur, simply for hiring. This is largely linked with the biggest indicator of quality of life, it being the infant mortality rate. Lower obviously being better. And for those vaccine deniers, it has been calculated that for _every dollar invested in vaccination, $16 is expected to be saved in healthcare costs and the lost wages and lost productivity caused by illness and death._

Now let’s go to human beings. We are way better equipped to longer runs than others. And here is why:

_The first advantage is in how we breathe. A quadruped can take only a single breath per locomotive cycle, because its chest must absorb the impact on the front limbs. We, however, can choose other ratios, and that lets us use energy more flexibly. The second (and greater) advantage is in our extraordinary ability to regulate our body temperature, which allows us to do what lions cannot: run long and hard in the noonday sun._

We went on into the entire field of motors and how important they are. Motors include e.g. 2 billion tiny 4mm diameter DC devices for making your cell phone vibrate. But these are also about the big ones and he postulated that _without the low operating costs, high efficiency, high reliability, and great durability of diesel engines, it would have been impossible to reach the extent of globalization that now defines the modern economy._ More on motors later.

He talked about Muybridge who build a device to film animals while running and in 1883 he _began an extensive series depicting animal and human locomotion. Its creation relied on 24 cameras fixed in parallel to a 36-meter-long track with two portable sets of 12 batteries at each end. The track had a marked background, and animals or people activated the shutters by breaking stretched strings._

The cool thing is that he invented that. And in 1878 Edison Patented the way to record and play back sounds.

_He designed a small device with a grooved cylinder overlaid with tinfoil that could easily receive and record the motions of the diaphragm. “I then shouted, ‘Mary had a little lamb,’ etc.,” Edison later recalled. “I adjusted the reproducer, and the machine reproduced it perfectly. I was never so taken aback in my life. Everybody was astonished. I was always afraid of things that worked the first time.”_

Another good things to think about: _And why do we measure the progress of economies by gross domestic product? GDP is simply the total annual value of all goods and services transacted in a country. It rises not only when lives get better and economies progress but also when bad things happen to people or to the environment. Higher alcohol sales, more driving under the influence, more accidents, more emergency-room admissions, more injuries, more people in jail — GDP goes up._

Let’s move into the biggest part though, power and energy. The thing is that a wind turbine probably needs a year to generate enough power to pay back for the environmental impact of the hundreds of tons of concrete and steel.

So, _for a long time to come — until all energies used to produce wind turbines and photovoltaic cells come from renewable energy sources — modern civilization will remain fundamentally dependent on fossil fuels._

And let’s take shipping: _What would it take to make an electric ship that can carry up to 18,000 TEUs, now a common intercontinental load? In a 31-day trip, most of today’s efficient diesel vessels burn 4,650 tons of fuel (low-quality residual oil or diesel), with each ton packing 42 gigajoules. That’s an energy density of about 11,700 watt-hours per kilogram, versus 300 Wh/kg for today’s lithium-ion batteries — a nearly 40-fold difference. […] The conclusion is obvious. To have an electric ship whose batteries and motors weighed no more than the fuel (about 5,000 tons) and the diesel engine (about 2,000 tons) in today’s large container vessels, we would need batteries with an energy density more than 10 times as high as today’s best Li-ion units. But that’s a tall order indeed: in the past 70 years, the energy density of the best commercial batteries hasn’t even quadrupled._

So where is all our energy usage going that we can hardly turn off:

_In contrast, several key economic sectors depend heavily on fossil fuels and we do not have any non-carbon alternatives that could replace them rapidly and on the requisite massive scales. These sectors include long-distance transportation (now almost totally reliant on aviation kerosene for jetliners, and diesel, bunker fuel, and liquefied natural gas for container, bulk, and tanker vessels); the production of more than a billion tons of primary iron (requiring coke made from coal for smelting iron ores in blast furnaces) and more than 4 billion tons of cement (made in massive rotating kilns fired by low-quality fossil fuels); the synthesis of nearly 200 million tons of ammonia and some 300 million tons of plastics (starting with compounds derived from natural gas and crude oil); and space heating (now dominated by natural gas)._

For example, _today’s jet fuel — the most common formulation of which is called Jet A-1 — has a number of advantages. It has a very high energy density, as it packs 42.8 megajoules into each kilogram (that is slightly less than gasoline but it can stay liquid down to –47°C), and it beats gasoline on cost, evaporative losses at high altitude, and risk of fire during handling._

The other big part are ammonia and liquid nitrogenous fertilizers … we need 150 million tons of NH4 a year, which produces 1% of the global greenhouse gas emissions. Of course I have to admit “have to” is a bit hard a term. But it comes to a good point: _In order to cut future nitrogen losses, we should do everything possible to improve the efficiency of fertilization, reduce food waste, and adopt moderate meat consumption . And even that will not eliminate all nitrogen losses — but that is the price we pay for having gone from 1.6 billion people in 1900 toward 10 billion by 2100. […] The UN’s Food and Agricultural Organization puts the annual global losses at 40–50 percent for root crops, fruits, and vegetables, 35 percent for fish, 30 percent for cereals, and 20 percent for oilseeds, meat, and dairy products. This means that, globally, at least one-third of all harvested food is wasted. […] Cutting food waste in half would lead the way to a more rational use of food worldwide, and the benefits could be huge: WRAP estimates that a dollar invested in food waste prevention has a 14-fold return in associated benefits._

But on for some interesting data:

We build short-stalked wheat because you know, the stalk takes nutrients to produce :)

_The average weight of American broilers rose from 1.1 kilograms in 1925 to nearly 2.7 in 2018, while the typical feeding span was cut from 112 days in 1925 to just 47 days in 2018._

On Wine: _By 1980 the per capita annual mean was down to about 95 liters a year, by 1990 it sank to 71 liters, and by the year 2000 it had fallen to just 58 liters, cut by half over the course of the 20th century. The current century has seen further declines, and the latest available data show the mean at just 40 liters a year, 70 per cent below the 1926 record. The wine consumption survey of 2015 (to be repeated in 2020) details deep gender and generational divides that explain the falling trend._

_The minimum water requirement per kilogram of boneless beef is, indeed, high, on the order of 15,000 liters, but only about half a liter of that ends up incorporated in the meat, with more than 99 percent being water needed for the growth of feed crops which eventually re-enters the atmosphere via evaporation and plant transpiration, and rains down._

_Bacteria account for about 90 percent of the human body’s living cells, and as much as 3 percent of its total weight._

And to finish:

_The affluent world has used hundreds of billions of tons of it to create its high quality of life, but right now we do not have any affordable non-carbon alternatives that could be rapidly deployed on mass scales in order to energize the production of enormous quantities of what I have called the four pillars of modern civilization — ammonia, steel, cement, and plastics — which will be needed in Africa and Asia in the decades to come._

My deepest respect go to the author, Adam Fisher.

So why is silicon valley special:

> Stone: The infrastructure is here: the real estate people, the legal people, the you-name-it people. They get start-ups, so it’s easier: “Oh, okay, you’re a start-up. So, here you go.” It’s just easier to do start-up stuff, because everyone in the whole ecosystem knows about start-ups.”

On what the early hackers were building with that internet:

> Jamie Zawinski: We weren’t building a toy. We were building a communications medium. We were letting people connect to each other in a way that they hadn’t been able to before. It was the opposite of television; it was giving people a voice.

The problem of the wrong people coming to silicon valley during the first bubble:

> Sean Parker: And then it becomes the post–social media era. It’s all the people who would have become investment bankers who want to go start internet companies, and it’s a purely commercial, purely transactional world. It’s just become this transactional thing, and it’s attracted the wrong type of people. It’s become a very toxic environment. A lot of people have shown up believing, maybe correctly, that they can cash in. But that’s Silicon Valley the ATM machine, not Silicon Valley the font of creativity and realization of your dreams.

Around the special kind of people that are founders:

> Steve Jobs: The people who really create things that change this industry are both the thinker and the doer in one person.

On building programming tools:

> Alan Kay: In programming there is a widespread theory that one shouldn’t build one’s own tools. This is true — an incredible amount of time and energy has gone down that rat hole. On the other hand, if you can build your own tools then you absolutely should, because the leverage that can be obtained can be incredible.

By the way, Flickr was a tool as part of a possible game and Slack an internal tool for a company building a game, both companies built by the same founders.

The really really and amazing story of the founding of Atari:

> Nolan Bushnell: I got on a flight to Chicago and called on Bally and said, “Would you like to license my next game? A driving game? It’s going to cost you this much of money in development.” They said, “Yeah, we’d love to do it!” And so I had my cash flow in hand. I went back to Nutting and said, “I’m going to have to leave.”

> Ted Dabney: We wound up getting this contract, $4,000 a month for six months to develop a video game and a pinball machine for Bally.

> Al Alcorn: Nolan then goes and hires me, employee number three, and we go off and start Atari.

Seriously? WTF!

> Ted Dabney: Bally paid for it. They paid $24,000 to us for that game, but Bally kept not accepting the game, kept not accepting it. Nolan and Al and I were sitting around looking at each other: What are we going to do?

> Al Alcorn: So Nolan said, “Put it out on location.” We put it next to a Computer Space at Andy Capp’s Tavern. I remember the day we put it in. Nolan and I popped it in one day after work and went and bought a beer and watched until somebody played it. I never thought anybody would play it! Think about it. There’s no instructions. It just says “Pong,” which meant nothing. There’s two knobs and a coin box. What’s the motivation here? So, some guy plays it. Nolan went up to him afterward: “What did you think of that?” And the guy said, “Oh yeah, I know the guys who made this machine.” And I’m thinking, Save the bullshit for the ladies. So we left. Not long afterward, Alcorn receives a phone call from Andy Capp’s Tavern.

> Al Alcorn: The machine stopped working. It didn’t surprise me. It was just thrown together quick and dirty. It was never meant to run on location. So, I went there after work and the attract mode was working, so that tells me most of the thing is working. So what’s wrong with it? I open up the coin box — it was a Laundromat bolt-on coin box — to give myself a free game to see what was going on, and when I opened that coin box up it was just jammed full of quarters. I can fix this! I told Nolan, “Wow, this is interesting…”

You stumble upon your greatest successes. Remember that.

> Michael Malone: It’s hard to capture just how crazy Nolan was in those early days. He was a wild man. He was young. He lived high. He had the Rolls-Royce. The code name for each new product was named after some hot girl on the assembly line. He really did have that master-of-the-universe thing. I mean there was coke with the assembly-line girls in the hot tub. This guy did the whole Cash McCall thing. He was just throwing out sparks in every direction.

or

> Ray Kassar: When I arrived there on the first day I was dressed in a business suit and tie and I met Nolan Bushnell. He had a T-shirt on. The T-shirt said: I LOVE TO FUCK. That was my introduction to Atari.

Nolan seems to have been a character.

Why corporates sometimes fail … :)

> Alvy Ray Smith: I get the call from Jerry Elkind, my boss. He says, “We’re going to let you go.” And I went, “Well, why?” And he said, “Well, we’ve decided not to do color.” I said, “But Mr. Elkind, the future is color. It’s obvious. And Xerox owns it completely!” He said, “That may be so, but it’s a corporate decision to go black-and-white.” Okay, bye.

Amazing stories on Steve Jobs:

> Al Alcorn: And he came back a few months later. I remember Ron Wayne came in and said, “Hey, Stevie is back.” And I said, “Steve who?” “Steve Jobs.” Oh that kid, yeah. “Oh, bring him in.” And I wish I had a camera. He was wearing a saffron robe, shaved head, barefoot, had a Baba Ram Dass book, Be Here Now. Steve gives me the Baba Ram Dass book and says, “Can I have my job back?” “Sure.”

Fun stuff you could do with the Apple II.

> Andy Hertzfeld: For the Apple II you could write programs that played music — but it had no real sound hardware. All you could do with the software was hit a memory address, which I still remember: C030. So if you hit C030 it would produce a “click” on the speaker. That was all the hardware could do. But if you did that a thousand times a second with software you’d get a one-kilohertz tone, if you did it three thousand times a second you’d get three kilohertz. And the Apple II was literally full of stuff like that, where by just using tiny little bits of resources it could do amazing stuff. In fact, just looking at the Apple II design gave you the feeling that anything was possible, if you were just clever enough. That’s what the main lesson.

But back to Atari, which is one of the best parts.

> David Crane: Pitfall took me about a thousand hours sitting at a computer to do. Now that’s a long time except it’s only a few months. Six or seven months. And it turns out that Pitfall earned Activision $50 million wholesale. So I made for them $50,000 an hour given the thousand hours I worked on it.

> […]

> Alan Kay: That year Atari’s gross just by itself was $3.2 billion, and that was several hundred thousand dollars more than the entire movie industry in Hollywood. So at that time Atari was bigger than all of Hollywood: They had money coming out of their ears!

> […]

> Alan Kay: They used to send the corporate jet up the coast to get shrimp for the executive dining room, and the joke was you could tell how Atari was doing by the size of the shrimp in the executive dining room.

> […]

> Howard Warshaw: By mid-1984 we were down to two hundred people. So, in about a year and a half the company goes from ten thousand employees to two hundred employees, and I was still one of those people. It was a dark time.

Simply Amazing.

> Jamis MacNiven: Titans rise, titans fall — that’s the nature of the world. It just happens faster in Silicon Valley.

The startup that was eBay and how Pierre came to his boss who didn’t see it.

> Michael Stern: So Pierre came to me when I was the general counsel in 1994 and said, “I’ve created this little electronic community. I’m getting people to talk to each other about trading tchotchkes. We’re creating traffic on the network and getting people into a community. That’s kind of in our sweet spot, isn’t it?” That was General Magic’s thing: the whole notion of electronic community. I said, “That’s the stupidest idea I’ve ever heard! If you want to do it, bye, see you later.” That was eBay.

And this was its power:

> Joey Anuff: When eBay came around I couldn’t think of anything better to do than buy every single toy my parents ever told me I couldn’t get. Like Bionic Bigfoot from The Six Million Dollar Man. Still in the box! It was a fucking miracle.

And there were funny side effects:

> Steve Westly: eBay has this stunning idea: that they will enable you to sell anything to anyone. But there’s these wild downsides, like, one Saturday I get a call from our engineering department saying the FBI is calling and they want to know why we have an operational rocket launcher for sale.

But let’s get to the eBay IPO:

> Jim Griffith: Goldman Sachs had scheduled the IPO for October of ’98.

> Steve Westly: We’re there with literally the most serious people from Goldman Sachs, and they’re saying like, “How big do you think the Beanie Baby playing-card sports-card trading-card industry is? One hundred million? Two hundred?” And that’s when I jumped up and said, “Well, you know we now have a thousand categories and we have an average eleven hundred items per category and we have a sell-through rate of X, and we’re adding categories at this rate, and by the way, we’re going into more geographic areas, and so we think this is a multi-billion-dollar company,” and we were able to convince the people from Goldman Sachs that we might be on to something.

And then came the first trade:

> Brad Handler: There was a long wait before the first trade. It was more than an hour after the market opened, and when the first trade came across the tape I think it was somewhere in the fifties. We all realized that our lives had all irrevocably changed. In that instant everyone in that room was probably worth more than their parents had ever made in their lifetimes.

This sounds amazing but at the same time, it was a website, with immense growth, so this is what happened there:

> Steve Westly: I was used to this, because I’d have to be handling marketing, and the press would start calling and asking when the site would come back up, and I’d talk to engineering and we’d say, “Fifteen minutes,” or “Forty-five minutes,” but that day there was an outage to the extent where engineering for the first time ever said, “We don’t know.” I said, “What do you mean, you don’t know?” This was a bad one.

> Pete Helme: It made the whole site unusable, and nothing worked, and people had to scramble, all the engineering teams got together, and Meg was there, rallying the troops: “How do we figure this out?”

> Jeff Skoll: Part of the problem was that the only person that knew how the system worked was Mike Wilson. Pierre had worked with him before at The Well. And Mike worked his tail off: He was in the office seven days a week. And Mike for the first time in three years took a vacation. He went off to the Caribbean somewhere and he was sort of out of touch for a week, and that was the week the system went down. Pierre had long since lost track of what was going on, Meg didn’t have a clue — that’s not her background — and neither did the other engineers in Mike’s group. The holy grail was held by Mike, so nobody really understood what was wrong.

> […]

> Pierre Omidyar: It was front-page news. We had CNN satellite trucks in the parking lot. It was big, big. It was, “The world is watching, this company is gone. It’s going away.”

> […]

> Maynard Webb: And Meg had promised, based on the June outage, that we would do a free listing day, which meant we took the fees away, and everybody would go crazy for twenty-four hours, and our site traffic would double or triple in volume — which is not necessarily a recipe for greatness when you’ve had scale issues.

It must have been fun, and hell, and fun, and hell, and fun.

> Steve Westly: The stock split twenty-four to one from the initial IPO. It turned out to be a thousand-to-one return on the venture investment. Still one of the biggest in history.

How Steve Jobs bought Pixar:

> Ralph Guggenheim: The story I was told was that there was a conference call between Steve and the execs at Lucasfilm, and you can imagine execs in a conference room with a speakerphone on the middle, and Steve’s on one end of the phone and all of these execs are trying to negotiate with him at the other end. Steve says something like, “I’ll give you five million dollars for the thing. I’m going to have to put in another five million dollars. That’s my offer!” And one of the execs undoubtedly said, “What! Five million? That’s not nearly enough.” And I was told that Steve said, “Fine, go fuck yourself!” and instead of going and fucking themselves they took the offer and they closed the deal.

And on to Steve finally finding out about the first Pixar movie, Toy Story:

> Alvy Ray Smith: So Steve’s busy running NeXT and Disney took the movie to New York and the critics saw it. They went nuts and they said, “This is going to be a huge success!” And as soon as Steve heard that he pushes Ed aside to be there when the cameras roll.

Then we go on to Netscape:

> Jim Clark: I clicked around, I found Marc’s e-mail, sent him an e-mail right there saying, “This is Jim Clark. You may not know who I am, but I am the founder of Silicon Graphics, and if you know me, you know that the news has come out that I am leaving Silicon Graphics. I would like to start a new company. I would like to know if you would like to get together and join me to talk about that.” Ten minutes later, he responded. Marc Andreessen: I said, “I know who you are.”

And how they thought about what to do:

> Marc Andreessen: We were talking one day and we said, “Well, this internet thing — no one takes it seriously but it’s growing vertically…”

> Jim Clark: All I had to do was look at the growth of Mosaic for a year and see that it grew to a million users, and I figured there was a network effect of people getting online.

> Marc Andreessen: The idea was to basically do a new version of Mosaic but from scratch, and in particular do it as a real product.

> Jim Clark: I said, “Okay, let’s hop on a plane and go out and recruit your buddies.”

> Aleks Totić: Then Marc calls: “Hey, we’re flying in — I think we’re doing it: We’re going to do the Mosaic killer.”

And Jim Clark about Jamie from further up, or simply about geeks:

> Jim Clark: He had half of his head shaved. Some stylistic statement on his part. I completely ignored it. It did not matter to me; I did not give a hoot. He was a great programmer, brilliant young guy. I do not think anyone bothered. People have just got to realize that a computer geek is respected on the basis of how much code he can write and the quality of code he can write. People do not give a crap what he looks like. If you generate good code quickly, no one cares. Jamie made everyone a lot of money.

And then a big one about Google.

> Scott Hassan: Very quickly, in six to eight weeks, we were able to build the whole structure of Google. It was mostly just Sergey and I from two a.m. to six a.m. in the morning.

And the drain Google took on the Stanford network.

> Larry Page: We caused the whole Stanford network to go down. For some significant amount of time nobody could log in to any computers at Stanford.

And that they won’t get broke:

> Andy Bechtolsheim: The question, of course, is, “How do you make money?” And the idea is, “Well, we’ll have these sponsored links and when you click on a link we’ll collect five cents.” And so I made this quick calculation in the back of my head: Okay, they are going to get a million clicks a day at five cents, that’s fifty thousand dollars a day — well, at least they won’t go broke.

Which in the end really was not so easy and if you read The Search you see the other side. But still they got first investment:

> Sergey Brin: He gave us a check for a hundred thousand dollars, which was pretty dramatic. The check was made out to “Google Inc.,” which didn’t exist at the time, which was a big problem. […] We hadn’t really discussed valuations and stuff like that. He figured it would pay off, and he was right. We finalized all the details on the round after that. I guess we figured if we didn’t agree later, that it would be a loan. He liked us and he just wanted to sort of push us forward.

Because the Google founders actually were at Burning Man:

> Brad Templeton: In the early years Sergey would just actually sleep in whatever camp he found himself in at the end of the night. Google is incorporated immediately after the return from Burning Man.

And it gets even funnier:

> Heather Cairns: They handed me a folder full of checks for like $100,000, $200,000 from Andy Bechtolsheim, Jeff Bezos, David Cheriton. They sat in the back of my car for weeks because I couldn’t get out of work in time to even get a bank account opened.

And at the end Google people just wanted to continue to be in a frat house.

> David Cheriton: I’m tempted to say it was like a frat house, although I’m not sure I’ve ever been to a frat house. It was a long way from a professional company — frankly, once I invested I really didn’t have the sense for some time whether these guys were really taking the company routine seriously or not, so I used to say, “Well, I spent $200,000 on a T-shirt.”

I mean check this:

> Charlie Ayers (note: first chef at Google): I come from the world of the Grateful Dead, so I know how to party, and so on the ski trips in Squaw Valley I would have these unsanctioned parties and finally the company was like, “All right, we’ll give Charlie what he wants.” And I created Charlie’s Den. I had live bands, deejays, and we bought truckloads of alcohol and a bunch of pot and made ganja goo balls. I remember people coming up to me and saying, “I’m hallucinating. What the fuck is in those?” It was like something out of The Hangover — hot girls drooling all over themselves and people passed out on the chairs.

> Larry Page: All sorts of fun things.

> Charlie Ayers: Larry and Sergey had like this gaggle of girls who were hot, and all become like their little harem of admins, I call them the L&S Harem, yes. All those girls are now different heads of departments in that company, years later. […] Sergey’s the Google playboy. He was known for getting his fingers caught in the cookie jar with employees that worked for the company in the masseuse room. He got around.[…] HR told me that Sergey’s response to it was, “Why not? They’re my employees.” But you don’t have employees for fucking! That’s not what the job is.

It was wild times back then.

> The joke used to be at the time, this being ’97 through 2000 — the golden era — was basically all you had to do was stand at the corner with a cocktail napkin, and VCs would throw money at you from a passing car. I loved it.

> Ev Williams: I remember leaving some party with a bottle of champagne. It was like, “You’re leaving? Have a bottle of champagne!” In those days, it wasn’t a ridiculous gift…

> […]

> Jamis MacNiven: There was a guy — the name of the company was Pixelon — he raised fifteen, eighteen million dollars, had no product, went to Las Vegas, hired the Who and, I think, Kiss.

> Jamis MacNiven: Draper Fisher Jurvetson once rented one of the biggest buildings in the world, Hangar One at Moffett Field — it was where the USS Macon zeppelin was — and filled it with amusement park rides. There must have been eleven thousand pounds of shrimp and it was raging. Barbara Ellison, Larry’s ex-wife, used to throw parties with zebras and elephants and giraffes. There was a party once that had army tanks driving around with naked women. One guy had a huge party and hired these four party girls, to just walk around completely naked, in Woodside, on a Sunday afternoon!

And you always think it is about work only over there. :)

> Ev Williams: The thing that people missed was that it was the financial bubble that burst, not the internet. This whole time usage of the internet continued to climb dramatically.

On to the next big one, Apple. Steve coming back wasn’t an accident.

> Larry Ellison: We used to go for these long walks in Castle Rock State Park over by the coast. It cost $5 billion to buy Apple in those days, it was run by a fellow named Gil Amelio. I had raised the money. I was going to give Steve 25 percent of the company, and we were going to take it private and he was going to run it. And Steve said, “You know, Larry, I think I figured out a way I can get control of Apple without you having to buy it.” I’ll never forget this conversation as long I live. I said “Okay…” and then he said, “Yeah, I think I can get them to buy NeXT Computer and I’ll go on the board, and eventually the board will recognize that I’m a better CEO than Gil Amelio and I’ll become CEO.”

Let’s get to the iPod:

> Tony Fadell: So I am getting on a ski lift in Vail, and I get this call: “Hi, this is Jon Rubinstein from Apple,” blah blah blah. The very next week I had my first meeting with Jon.

> Jon Rubinstein: There was a key meeting. It’s me and Phil Schiller and Jony — there was a group of us, but Tony is the lead.

> Tony Fadell: Steve comes in the room, does not say a word, takes the checklist, throws it on the desk, and gives the vision for it. Ron Johnson: It was, “I want to make a music device and I want it to hold all your music, I want it to be digital, with great software so you could take your music everywhere.” I don’t think he had thought of “a thousand songs in your pocket” yet, but that was what he was imagining.

How Apple used being small back then:

> Jon Rubinstein: We put together a pitch to the music companies: “Look, we’re at 2 percent market share with the Mac — we are not successful, so you have no downside to licensing us the music. You can do this experiment and it will cost you nothing.” If we’d been wildly successful in the market, there’s no way they would have licensed it to us. We didn’t matter, so they could license to us and who cares, right?”

> […]

> Mike Slade: People didn’t really think that the iPod was going to be a viable product for Windows, because it was a pain in the butt to hook it up to FireWire. It was kind of a hack for a Windows user until the USB 2.0 iPod came out. That’s one of the reasons why the music companies were willing to give him the Windows option. Jon Rubinstein: By then it was, “You’re kind of pregnant, so it is too late to back out now.”

Then comes facebook and its first investment:

> Mark Pincus: And then Sean put together an investment round quickly, and he had advised Zuck to, I think, take $500,000 from Peter Thiel, and then $38,000 each from me and Reid Hoffman. Because we were basically the only other people doing anything in social networking. It was a very, very small little club at the time.

And Facebook worked at night:

> Ezra Callahan: When we first moved in, the office door had this lock we couldn’t figure out, but the door would automatically unlock at nine a.m. every morning. I was the guy that had to get to the office by nine to make sure nobody walked in and just stole everything, because no one else was going to get there before noon. All the Facebook guys are basically nocturnal.

And they actually pushed code in the middle of the night to not effect too many people.

> Ruchi Sanghvi: So then we would only push out code in the middle of the night, and that’s because if we broke things it wouldn’t impact that many people. But it was terrible because we were up until like three or four a.m. every night, because the act of pushing just took everybody who had committed any code to be present in case anything broke.

DOMINATION!

> Ezra Callahan: We had company parties all the time, and for a period in 2005, all Mark’s toasts at the company parties would end with “Domination!”

> Mark Zuckerberg: Domination!!

> […]

> Max Kelly: We literally tore the Yahoo offer up and stomped on it as a company! We were like, “Fuck those guys, we are going to own them!” That was some malice-ass bullshit.

> Mark Zuckerberg: Domination!!!

> […]

> Steve Jobs: I admire Mark Zuckerberg. I only know him a little bit, but I admire him for not selling out — for wanting to make a company. I admire that a lot.

So what is the difference between Google and Facebook? (and then on to Odeo and Twitter)

> Ev Williams: Roughly speaking, there are two engineering cultures in Silicon Valley, which you could describe as hackers and engineers. And obviously Google is engineers. Engineers generally have computer science degrees and higher. They study the fundamentals. Hackers just want to make stuff work, and it’s not about doing it right necessarily. Facebook was kind of famously built by hackers, but they’re not the hippie hackers. Hippie hackers are a particular strain of hackers, and Noah is a total hippie. And so we hired hippie hackers — I mean Noah really hired these people. They are his people, he got along with them.

> […]

> Biz Stone: Those guys were a pain in the ass. They would specifically sit down during a stand-up meeting — on purpose! There was one stand-up meeting where they were sitting down and Ev was like, “Guys, please, I need everyone here by ten a.m.” Because everyone was showing up at noon or whatever. And one of the guys raised his hands and said, “I have a question.” And Ev was like, “Yes?” And he said, “What’s our motivation?” And Ev just lost it. Ev just yelled out, “It’s your fucking job!!”

I feel Ev. :)

So the launch of Twitter at South by Southwest.

> Ray McClure: Twitter just totally took over South by Southwest that year.

> Biz Stone: A couple of things that happened. One is this story of the guy at the bar who tweeted, “This bar is too loud. If anyone wants to talk about projects let’s go to this other bar that’s quiet.” And then in the eight minutes it took him to walk there eight hundred people had showed up! That was my realization that, Oh, shit. We definitely invented a new thing. We definitely made something that the world had never seen.

This was a great book to read. So let’s end with this:

> Nick Bilton: And at the end of the day, Twitter became a place for people to become more famous, and not a place for conversation.

> @RealDonaldTrump: My daughter Ivanka thinks I should run for President. Maybe I should listen.

> Nick Bilton: Twitter today is one person talking to a lot of people — not a conversation. And I don’t think that we as human beings were designed to enable 320 million people to have a conversation together.

> Steven Johnson: It’s true, Twitter is not a conversational medium, and that’s fine. Not everything has to be a conversation medium. So if you are judging it by that standard, then yes, it’s a failure. But I have found Twitter, personally, to be just a great addition to my life.

> @RealDonaldTrump: I love Twitter… It’s like owning your own newspaper — without the losses.

It’s written by John LeFevre who has started it all with registering a Twitter account called GSElevator, an account built by telling the tales out of the Goldman Sachs Elevator.

> Everyone else doesn’t get it: the lifestyle, the long hours and canceled plans, the binge drinking, and the nihilistic sense of humor. I don’t think any of them have ever woken up from a blackout on a subway car and gone direct to office still wearing last night’s tuxedo, only to be showered with praise. How could they understand? And they certainly don’t have the disposable income — although it doesn’t always feel limitless at the rate I spend it.

Even if the stories are only 50% true, they are so over the top, it is mind boggling.

> Banking has given me a very distorted view of money, priorities, and what I deserve or feel that I am entitled to. Despite my paycheck, I feel poor. This detached sense of reality, combined with my increasingly selective social circles and a blindingly aspirational culture, has created a lifestyle as precarious as any middle-class one. The numbers are just bigger.

The money that is flowing there is ludicrous and the way they are just pushing deals down the pipe because they need to be winning against the other bank, tells you that this entire market is somehow wrong. At the same time, their services are needed.

And John is doing a good job in walking that tight rope between entertaining over the top stories …

> I realized a very important thing that had eluded me during my exploratory trip a few months prior: that Lipton was leaving Hong Kong, not because of the opportunity in New York, but because he had a coke problem and his wife was threatening to divorce him. It’s a good thing that I wasn’t moving here with a wife or a coke problem. [ … ] Here I am, locked in a stranger’s bathroom, staring at myself in the mirror, sweat pouring down my face, white powder all over my nose, and my limp cock in my hand. I’ve been in Asia for all of eight weeks.

… and showing what it is like to work in international banking, and how the processes work, banks interact, clients demand, real people work on hundreds of millions of dollar deals.

Entertaining book.