Patrick Mathieson

software investor @tobacapital

[Patrick Mathieson @pmthn]

If humans and AI are substitutes, the asset-light vs. asset-heavy divide is sus, by @pmthn https://t.co/CtZotam9VX

[Patrick Mathieson @pmthn]

software companies earnings skyrocket b/c their 2nd largest cost center deflated massively

[David Orr @orrdavid]

@neeravbm What happens when the cost of making those features is 99% less?

[Patrick Mathieson @pmthn]

@taobanker oooh i like this

[Patrick Mathieson @pmthn]

@ser_kingslayer_ @evrgn11112231 "left for dead" they're up 41% and 56% YoY???

[Patrick Mathieson @pmthn]

@ShanuMathew93 it will more than 5x in 2 years

[Patrick Mathieson @pmthn]

I don't recall any prior technologies so compelling that people were publicly begging $1T companies to absorb their personal health data in order to make use of it

[Fabricated Knowledge @fabknowledge]

When Fable works it’s brilliant, but the unilateral guardrails makes me frustrated beyond belief.

I have a folder of health information for my fiancée with like 100 days of oura health data, a hundred lab tests, transcripts of doctor visits and like way more for a super detailed health file to help figure out a complex fatigue / chronic illness she’s been going through. Guess what, fable considers it unsafe. Why? F do I know, I was really excited to try to clean up the folder and make some better protocols but guess what, it’s unsafe to help her health.

There’s a private investment in the life science tools space, guess what it’s not safe. I’m trying to do some code scanning for vulnerabilities, not safe. I get the safety but it feels so incredibly out of touch for a group of a few 1000 all making total comp in the millions telling me what is and isn’t safe.

If Dario is worried about inequality I think he has to realize he IS the inequality, and the unilateral gatekeeping feels whack as hell. I don’t like it, and yeah I’ll try to keep using fable, but OAI can’t catch up fast enough.

[Patrick Mathieson @pmthn]

uhhhhhhhhhhmmmm

[Taelin @VictorTaelin]

this is my personal singularity moment

this post may sound like a paid ad. I only wish. I'm concerned, more so than happy. the world is changing, and, among the scenarios where AI goes terribly wrong, inequality is the most realistic, yet, the one Anthropic seems to be the least concerned about. I'm glad OpenAI is taking the opposite stance: *personal AGI for everyone*. I think this is a commendable position in the times we live. but who am I in the queue of the bread?

anyway, Fable is here, so I'll just report my first-hour experience

first of all, all my pet prompts are solved.
→ λ-calculus puzzles
→ bug questions
→ one-shot apps
all are trivial to it.

I don't have anything harder other than my
ongoing work

so, in the last several days, I've been toying with HVM5, a new interaction net evaluator with a faster loop.

after writing the first version, I left 32 GPT-5 agents working for ~20 hours each. this resulted in up to 2x speedups, but the file size increased by 2-fold and quality decreased significantly.

I then simplified the whole thing into an even simpler core, and left Opus 4.8 and GPT 5.5 optimizing it for 8 hours. Opus got a legit 6% - 34% speedup in most benches. GPT got better results, but, sadly, an unusable file.

I then asked Fable to optimize it.

2 hours later, it landed a 1770% speedup in one case, 100%+ in other 4, and 22% in average. yes, in 2 hours it outperformed me, opus 4.8 and a swarm of gpt 5.5 agents, by one order of magnitude.

that could not possibly be legit. "it must be hardcoding the benchmarks" (GPT trauma). so I read its explanation and what it did was, indeed, the most high impact optimization one could try first. seems like HVM5 was wasting a lot of time garbage-collecting unused branches of pattern-match nodes. I had optimized that for static mats, but not for dynamic mats. skill issue. Fable figured how to do it for these, resulting in a massive speedup in some benches

but wait, is that *correct*? I'm not sure yet, it is credible, but this is the kind of thing that is very easy to get wrong on interaction nets. the problem is, when I was ready to start auditing Fable's solution so I could tell whether it was buggy or legit, it interrupted me to tell me it had found a massive bug on the code *I* had written.

... wait, what?

so... for garbage collection purposes, I stored a bit on lambda term pointers that meant "the variable bound by this lambda has been freed, so, its lambda must free whatever argument it is applied to". that's fine. yet, on duplicator nodes, I also used the same bit to mean "one of the duplicated variables was freed, so, treat this dup as a passthrough no-op". so, if a lambda entered a duplicator, it would mistake the lambda's collection bit for its own, resulting in corrupted interaction!

that's a mouthful, why I'm writing this?

just so you can appreciate the sheer absurdity of what just happened. I didn't ask it to find bugs. I asked it for an optimization. and even if I did ask it to find bugs, this bug is so astonishingly subtle and specific, identifying it takes mastering the domain to an extent that it beyond even me. I'd easily need hours or days to fix it, *if* I ever came across it. chances are it would just go unnoticed. and Fable found it and fixed it like it was nothing, while it was busy adding a 17x speedup to a file that neither I, nor Opus 4.8, nor a fleet of GPT 5.5 managed to barely make 2x faster.

oh and there is also another tab where it is also ripping through Bend's codebase and finishing everything I had to do

I don't know what to say anymore

this isn't about Anthropic or OpenAI, this is about our collective future as a species. the world is changing, and we need to be aware of it, and discuss how to handle this change.

receipt below . . .

[Patrick Mathieson @pmthn]

@RHouseResearch pretty sure GOOG is doing all of the above, hence the equity raise

[Jake Bolling @jakebolling]

Today I’m thrilled to announce Scotch has raised a $20M Series A, led by VMG Partners, with participation from @firstround , @LererHippeau , and @TobaCapital .

In 2024, we set out to fix something that's been broken for 30 years: the technology running America's 40,000+ independent liquor stores. Most of those stores operate on POS software built before the iPhone. Before Amazon was founded. In some cases, before the store owners themselves were born.

Since launching our first store ten months ago, we've crossed $1 billion in annualized gross payment volume. That number tells us two things: (1) The problem is real, and (2) operators are ready to move.

Store owners are quite literally blowing up systems they've used for two decades to partner with Scotch.

While these milestones are fun to celebrate, they're far from what we're focused on. We think about the calls from owners and GMs who used to spend Monday mornings buried in distributor invoices, line by line, for hours. Now they spend that time looking at margins, planning reorders, and growing their store. That's what we built this for.

Our CTO, Dan Chen, spent more than a decade in liquor tech, including CTO at Drizly before its acquisition by Uber. Kevin Hodges and I built Skupos in convenience retail. We didn't stumble into this category. We picked it on purpose, and we built the team to win it.

This round lets us keep building faster and with more intention. More automation. More time returned to the amazing people running these stores.

Independent liquor retail is an $80B market that technology has ignored for a generation. Tens of thousands of stores will be getting the technology they deserve.

We're just getting started.
https://t.co/lYkGI7tK1b

[Patrick Mathieson @pmthn]

Berkshire is now another way to indirectly own Anthropic.....

[Dennis Hong @dennisihong]

Just having fun with LLMs:

Could Berkshire be the dark horse of AI infrastructure?

If Berkshire chose to enter the foundational compute layer (data centers, power, land, cooling, transmission), they’d bring:

Power
•Berkshire Hathaway Energy controls regulated utilities across the U.S.
•They own generation, transmission lines, and land, all increasingly critical bottlenecks for hyperscale AI clusters.
•AI datacenters are essentially electrical utilities with servers attached — Berkshire already runs those.

Capital
•Microsoft, Amazon, Google, Meta are all capital constrained by shareholder expectations.
•Berkshire is capital unconstrained.
•AI infrastructure is a $1T+ capex cycle, and BRK could underwrite multi-decade returns comfortably.

Connections
•Deep government relationships (federal + state) via utilities and insurance
•Long-standing ties with the largest CEOs in the world
•Trust and predictability that hyperscalers need when building 20+ year energy infrastructure

Berkshire could play the role of neutral, deeply capitalized sovereign-like partner for cloud/AI companies — especially as AI energy demand has become a national priority.

[Patrick Mathieson @pmthn]

$gtlb

cc @midwit_capital

[Patrick Mathieson @pmthn]

@Rthomas731 @agraybee also why is there gravity on the asteroid

[Patrick Mathieson @pmthn]

@evrgn11112231 I continue to be surprised at how many panicked negative takes there are on twtr regarding $META's performance. So it's -10% L12M, who cares? Does anyone on this app have a time horizon longer than a couple of quarters...?

[Patrick Mathieson @pmthn]

@SethLiebert @JaredSleeper because eventually MU will have a face ripping drawdown and everybody giga long memory will wish they were holding CRM

[Patrick Mathieson @pmthn]

@JaredSleeper indeed but it also seems like the q1 saas collapse was caused by fund managers who used claude code exactly once and then got short software

[Patrick Mathieson @pmthn]

good takes

[Jared Sleeper @JaredSleeper]

Maxi/mini takes on AI:

Maxi-

1) Cybersecurity will be an enormous consumer of tokens. All public-facing code needs regular, in-depth pen testing.

2) We are barely scratching the surface in code. Enterprises will have tens of thousands of apps, dynamically generated apps, per user/per customer apps, etc.

3) Enterprise adoption is still nascent. Leading companies are spending 10x+ as much on tokens as laggard companies.

4) Video/image understanding/generation will be huge drivers of token spend- as will anything to do with operating IRL/physics.

Mini-

1) Many use cases (consumer chat, various enterprise "agent" use cases) are close to intelligence-saturated. Intelligence saturation will render many major use cases today dirt cheap in a few years. See @StatueofIBBertY's posts on this topic.

2) Enterprise spend rationalization is real and imposed spend ceilings may be durable. As costs decrease/efficiency increases token volumes will go up, but budgets may not.

3) The global compute "shortage" is driven by capacity hoarding, not inference volumes per se (yes individual companies are inference compute constrained, to be clear). "Enterprise" adoption probably won't be fast enough to keep up with capex, we need prosumer-y use cases like chat/code that are very responsive to model quality increases. If we go 8-12 months without those, funding for training cools even slightly and hyperscalers like Meta pivot to being neoclouds, we could easily find ourselves in a compute glut.

4) No one knows what % of frontier model API spending is from cleverly hidden distillation attacks (or partly motivated thereby). Many, many deep-pocketed actors are motivated to do this at scale.

In general, I don't know if there's ever been a cycle where there are so many considerations on both sides that have order-of-magnitude-sized impacts on how supply/demand work out for tokens, infra, etc.

What a time to be alive.

[Patrick Mathieson @pmthn]

SaaSpocalypse canceled

[Patrick Mathieson @pmthn]

wtf?

[No Lie with Brian Tyler Cohen @NoLieWithBTC]

Trump’s DHS Secretary Markwayne Mullin just announced that they are “drawing up plans” to block all international flights into blue cities like New York, Los Angeles, and Chicago.

[Patrick Mathieson @pmthn]

"Indeed, I still haven’t heard a compelling explanation why all the terminal value question for public software companies isn’t doubly applicable for private AI-native software companies." https://t.co/jqxSx0CY4I @borrowed_ideas

[Patrick Mathieson @pmthn]

if AI automates low-level rote tasks, but human high-level judgment and taste is safe

then why is every CEO running strategic planning exercises 24/7 with Claude

and almost nobody can get customer service bots to work?

[Polymarket @Polymarket]

JUST IN: Uber’s COO says heavy AI spending is getting harder to justify, as higher token usage fails to show a clear payoff in consumer features.