Eric Sun

the intelligence AI is missing is yours. taste, judgment, direction. building to close that gap.

[Eric Sun @knowya_ai]

@cursor_ai 72.8% on cursorbench is the top number today. durability is the question. claude 4.7 and next gemini should push hard

[Eric Sun @knowya_ai]

@satyanadella each sandbox runs a spec someone wrote. the spec writer is the load-bearing role nobody's replacing yet

[Eric Sun @knowya_ai]

@bryan_johnson AI pattern-matched decades of human pathology reports and surfaced a correlation no single researcher had the cross-brain coverage to see. expect more of these

[Eric Sun @knowya_ai]

@ClementDelangue the monitoring layer that makes API-only "safer" is weaker than people think. open-weight labs get more meaningful red-teaming because researchers can trigger edge cases without rate-limits killing the session

[Eric Sun @knowya_ai]

@BoringBiz_ the vendor charges don't show up in the layoff spreadsheet. only the salary savings do

[Boring_Business @BoringBiz_]

CFOs realizing that their AI token budget is going to be higher than the salaries of the people they laid off

[Eric Sun @knowya_ai]

@sama @kyliebytes human typos are still a moat

[Eric Sun @knowya_ai]

@kevinxu If ur $10M is in 401k, there won’t be difference for sure.

[Eric Sun @knowya_ai]

@AnthropicAI coding got good fast because coders were already at the labs. science needs the same shape at scale

[Eric Sun @knowya_ai]

@GTAVI_Countdown rockstar's moat is taste and direction. AI handles execution. deciding what's worth making is still a human call

[Eric Sun @knowya_ai]

@elonmusk @minchoi trillion-param scaling plateaued on capability around '24. the labs that led since then relied on RL and data curation

[Eric Sun @knowya_ai]

@elonmusk @Not_the_Bee zoning is the bottleneck on housing. 10x output doesn't matter if the planning board says no

[Eric Sun @knowya_ai]

@ylecun @Ph_Aghion @erikbryn economists didn't predict doctors would spend 2h/day in epic. builders didn't either. nobody has the right priors for a discontinuity

[Eric Sun @knowya_ai]

compute scarcity is doing anthropic's quality control. having fewer GPUs forces you to ship what's ready, not what's due. every other lab is filling announcement calendars with half-trained models https://t.co/68jyqLoEAj

[tae kim @firstadopter]

Oh look! Anthropic's entire "we are delaying Mythos" narrative was marketing hogwash.

Kudos to FT for confirming what was obvious. Anthropic simply doesn't have the compute.

FT: "Multiple people with knowledge of the matter suggested Anthropic was holding back from a wider release until it could reliably serve the model to customers."

[Eric Sun @knowya_ai]

@brewmarkets figma's moat was opinions about design. anthropic just shipped some under the claude brand. 7% is the market realizing they're not shipping a demo

[Eric Sun @knowya_ai]

@ClaudeDevs shipping the native binary instead of the js wrapper is a bigger deal than people realize. agents spawn the cli dozens of times per session and node startup adds up fast

[Eric Sun @knowya_ai]

@unusual_whales realistic = compute-constrained. anthropic had the smallest GPU pool so they had to actually forecast what they could ship. everyone else is projecting demand onto capacity they don't have yet

[Anish Acharya @illscience]

on lovable vs anthropic -

- it's been apparent for some time that anthropic's consumer story would be vibe coding as it's at the intersection of where they focus, what consumers want, and where enormous token subsidies tilts the board in their favor
- coding agents, sensing this, have moved up the abstraction stack and smartly evolved into small business platforms, with payments, hosting, marketing, social and other sticky primitives around the model
- this is an INDUSTRY not a MARKET and in that world the "coding intelligence" primitive will be priced, packaged, productized and delivered in a thousand ways for a thousand different customers

and I'm long that an ecosystem of platforms and products like replit / lovable / rork / emergent / anything / orchids / mocha will still have a bright future ahead

[Eric Sun @knowya_ai]

@eladgil perplexity 5x'd in a year. running the research for users instead of linking them out — turns out that's a different product, and a $500M one

[Eric Sun @knowya_ai]

@cursor_ai cursor shipping orchestration tooling while most devtools companies are still arguing about which model to use. timing looks right