epochso

“We were in the jungle. There were too many of us. We had access to too much money, too much equipment, and little by little, we went insane.” replace equipment with information and this quote perfectly describes the information age

- capital/asset intensity has nothing to do with where value gets captured - model layer has a much more favorable market structure than app layer (concentrated with high barriers vs crowded with no barriers) - the company at the bottom of the value stack tends to get most of the value. OS > app

Someone please tell me why all the value in AI won’t accrue at the application layer vs model layer > application layer has minimal capex spend vs model layer spending billions in capex > asset light, variable cost structure business which makes it easier to react to macro shocks > lower leverage and debt burden because you didn’t have to raise billions in debt for GPUs and data centers > application layer can focus on a specific vertical and go all in on catering to that customer base while model layer is more generalist > intense competition at the model layer (OpenAI, Anthropic, Google etc) while application layer still remains open for the taking. TAM feels much larger for applications because there are so many verticals you can chase What am I missing?

@SuperMugatu This. Inflationary in the short run but deflationary in the long run.

A news reporter asked Michael Jordan if he thought the ’90s Bulls could beat LeBron’s Lakers. MJ: Yes. Reporter: By how much? MJ: Two or three points. Reporter: Why so close? MJ: Most of us are almost 60 now.

Everyone in finance should be thinking long and hard about this. There’s fewer seats at the top and not everyone makes MD or PM.

@evrgn11112231 Everyone’s a momentum investor

The overlap between funds gigalong semis today and net cash / underexposed semis/tech for most of 23-25 is shockingly high.

My Apple Car.

@ContrarianCurse Generally want to be long the seven sins

Every time $DASH grows GMV people are like omggg people are soooo stupid for using this First of all, if a home made meal is $20 and DoorDash is 35, still might be the cheapest way I can buy back an hour with kids Second, if you use it more than twice a month, spend the $10 (if that, there has been free ways to get it) on the dash Third, literally every day 200 restaurants have bogos and various deals that make it cheaper than if you went there. The reason is because on excess capacity as restaurants the incremental is very high so they’ll promo until they can’t serve anymore Like basically any food you can imagine at the palm of your hand for cheaper than groceries most of the time and people can’t fathom why the growth doesn’t stop

@soleio Network economies may be the last moat standing as switching costs approach zero

I’m too Facebook-brained to have an unbiased take here, but I don’t understand how more software isn’t built around recurring networks of people as core infrastructure. We barely get multiplayer right in most software experiences!

The NBA has become unwatchable garbage. What used to be a sport of skill and athleticism has become a sport of twinkle toes traveling and flopping. Corrupt refs aren’t making things any better either

@TakoTreba @Airbnb This happens all the time and the company does nothing about it.

On Sunday evening, I had the worst experience I’ve ever had with @Airbnb. A few hours before arriving in Hamburg for OMR, a conference that brings more than 70,000 people to the city, my booking was suddenly cancelled. I immediately contacted customer support, expecting them to help me find another place. Instead, they told me they couldn’t assist because “Airbnb didn’t cancel your booking, the host did.” Excuse me? This happened around 11 PM. I spent the entire night on calls with customer support and did not sleep at all. At around 1 AM, things got even worse. I received a message saying I was no longer allowed to make bookings on Airbnb because my account had been flagged as potentially dangerous. I honestly couldn’t believe what was happening. I was traveling with my three-month-old baby. I explained that multiple times. It made no difference. No meaningful help. No travel credits. My account was blocked. And by the time I arrived in Hamburg, I had to scramble to find a hotel on my own. The ironic part? There was no AI involved in this experience. Only humans following scripts and passing me from one person to another. And, Brian, I genuinely wish I had been dealing with AI instead. It would have been faster, more logical, and probably far more empathetic than the people I spoke with that night. What shocked me most was realizing how few real alternatives there are to Airbnb. At midnight I was completely on my own trying to solve a problem that should never have happened in the first place. Worst customer experience I’ve had in years.

Cormac was on a different level man

@evrgn11112231 @malinvested True. Although hyperscalers doing just fine turning tokens into profit and Google has plenty of internal use of the compute. I don’t think MSL is a read through for anything other than Zuck’s willingness to bet the farm when he thinks his terminal value is at risk.

They are the only ones that actually build on the compute they buy, everyone else is in the merchant biz and their customers are like meta (or are meta as well see anthropic) but are in worse position to monetize. So if you think meta is wasting money and can’t monetize AI the whole thing goes up in smoke because no one is better positioned to turn tokens into money than meta imo.

"no, but this time there's earnings support. this time the fundamentals really are improving. there's real demand. back then it was all fake. pets dot com bro." don't know if this is a "bubble". not really trying to make that call. but please stop, implicitly or explicitly, comparing companies like pets dot com to those like SanDisk. pets dot com was a ~$400mm market cap at peak, call it ~$700mm in today's dollars. today, we have plenty of names with market caps in the multiple billions that are burning cash with negligible revenue and highly uncertain prospects of ever generating any meaningful revenue. you want pets dot com, we got pets dot com aplenty. that doesn't mean the whole thing's a bubble. but if that's the litmus test you're gonna go with: ✅ the intellectually honest comparison is something like the names from the below table. SanDisk is nice because it happens to be in both lists (different industry structure, end markets, etc., sure, not really the point). then (2000 / 1999) vs. now (2026E / 2025): revenue growth: 2.4x (then) vs. 3.9x (now) GP: 2.6x vs. ~9x EBIT: 4.1x vs. ~18x 1-yr stock price performance at mkt top then: 10x, current 1-yr performance: 40x stock price performance relative to fundamental performance tracks pretty close so by all means make an argument about why it's more sustainable this time, but please don't pets dot com us

@evrgn11112231 @malinvested Tbf this says more about META than the industry. Metaverse 2.0

@malinvested It’s also funny bc today you have one of the biggest capex spenders ($META) explicitly telling you they may actually be overbuilding and have no idea yet and we’ll see soon and if it turns out they are they will cut back spending.

@JosephJacks_ 😂

PREDICTION: Anthropic will surpass Alphabet in revenue by mid-2028. This is not a bull case or an acceleration scenario — it is a continuation of the curve already in evidence. Anthropic’s ARR went from $1B (Jan 2025) to $9B (Dec 2025) to $30B (Apr 2026) — a 3.3x step in a single four-month window, and the curve has been steepening, not flattening. My projection actually assumes deceleration from here: $100B by end of 2026, $340B in 2027, $850B in 2028, $1.4T in 2029, $2T by 2030. Crossover with Alphabet happens at ~$575B in mid-2028, not because Anthropic accelerates beyond today’s pace, but because Alphabet — locked at ~15% YoY in a mature ads-and-cloud business — cannot match enterprise AI’s adoption physics. As @rodriscoll intelligently observed recently, Gemini tokens served grew by only 60% in the last quarter … while Anthropic grew by 10X. Three drivers make the continuation structural, not speculative: customers spending >$1M/year with Anthropic doubled from 500 to 1,000 in under two months post-Series G (these are multi-year expanding contracts with near-zero churn — switching a deployed agent stack mid-flight is operationally untenable); Claude Code is the wedge, not the product, dragging the rest of the platform — agents, MCP, healthcare, biotech — into every Fortune 2000 deployment as an attach point; and compute supply is finally non-binding with the 3.5GW Google + Broadcom deal (2027+), this weeks SpaceX partnership, and 1GW of standing Google capacity for 2026. For most of 2024–2025 the bottleneck was supply, not demand. That constraint is releasing exactly when the demand curve is steepest. The standard objection — “no company has ever sustained this at scale” — applies a software-era frame to a labor-era business. AWS, Azure, and Meta decelerated at $50–100B because they sold tools to the economy. Anthropic is selling cognitive capacity into the economy. The TAM isn’t enterprise software ($800B). It’s labor ($50T+). When the denominator is two orders of magnitude larger, “deceleration at $100B ARR” stops being a law and starts being an assumption. The crossover isn’t a maybe. It’s a function of timing. Mid-2028 is when I think Anthropic surpasses Google.

"AI code is crap." The shit your human engineers get up to:

@johnarnold It's rational behavior in a world driven by power laws. If only 4% of stocks drive all the returns, the reward for correctly identifying the 4% is so massive that it economically justifies the pursuit, even if the "average" picker fails.

There’s a reasonable argument that small investors were better off before finance was democratized, when a stock trade cost $20 and research wasn't easy. Those frictions discouraged active trading in favor of 'buy and hold' investing, a far better strategy for the average person.

@pmje73 @fundiescapital All software got priced as if it were a great business, but most are mediocre — the absence of real economic profits makes that obvious. To Paul’s point, all it took was one reasonable question about the long term to move the market from EV/S to P/GAAP EPS.

Yes but we also saw terminal multiples assume one bullet proof view of the future and now see terminal multiples reflect a more realistic uncertain future. Even if fundamentals don’t collapse the perception of the terminal should be lower, at a minimum to force an efficient use of capital and greater respect for dilution. My comment might be viewed as an expectation of a negative outcome when really it’s meant to highlight the reality of an uncertain one.

A few years ago I mentioned on a podcast that “biz quality” is overrated as an analytical tool and industry structure/barriers to enter the industry are more important. A simple case study of the fragmented SW industry vs the oligopoly like semi and components sub sectors shows how these industry structures can just sit there for years waiting for an internal or external catalyst to come along and expose those structural weaknesses or strengths.

@pmje73 I call this Buffett’s sin. He convinced generations of investors to invest by assessing biz quality. But that’s not the game in a pari mutual betting system.

san francisco in one picture

See the announcement today from @brian_armstrong at Coinbase in the attached pic. Managing fleets of agents effectively is the most high-impact thing you can do now for your company and for yourself personally. I have the most advanced set of open-source, provider-agnostic agent tooling to do all that at scale: MCP Agent Mail for inter-agent messaging and file reservations (avoid the git worktree anti-pattern and merge hell). beads_rust for agent-intuitive task management that works well across all harnesses. bv for triaging tasks based on the graph structure of task dependencies (maximize development velocity through de-bottlenecking). ntm for doing agent orchestration (creating and managing swarms of Claude Code and Codex instances). dcg for preventing the agents from doing destructive things. cass for letting agents instantly search all session history globally for any agent harness ubs for polyglot “super linting” and bug finding. You can find all of these on my GitHub: github.com/Dicklesworthst… Or install all of them automatically on a Linux cloud server using the acfs setup wizard, available for free at agent-flywheel.com And see my complete guide explaining how to do planning with these tools, which is the key to achieving high code quality: agent-flywheel.com/complete-guide And then to make these tools as effective as possible, I have a complete suite of powerful and sophisticated agent skills that are deeply integrated with the tooling available at jeffreys-skills.md The skills are $20/month. You don’t need to use them but they’re absolutely worth it if you want to get the most out of the tools and the agents in general. Why do this now? What’s happening at Coinbase is going to happen soon at every well-managed technology company within the next year. The writing is on the wall. This isn’t alarmist fear mongering. This is the inexorable reasoning of the market and capitalism grappling with these irresistible economic forces. Don’t wait and become a statistic, facing a structurally difficult job market without the skills that all well-positioned employers are going to be looking for. Take action NOW to master these technologies and empower yourself.