I’m joining Coinbase (biggest crypto company in the US) as a Tech Lead. There are countless phenomenal big tech and startups out in the industry right now, and it’s always so difficult saying “next time” to other amazing opportunities. #NFT #Coinbase @tezos

GPT-5.2 solves our COLT 2022 open problem: “Running Time Complexity of Accelerated L1-Regularized PageRank” using a standard accelerated gradient algorithm and a complementarity margin assumption. Link to the open problem: https://t.co/JcdTbhnLky All proofs were generated by

but jobs? 85m gone, 97m new by '25; bias widening gaps. risks? - existential (bio-threats), - cyber (prompt hacks), - bubbles—vc hyper-focused, four ai stocks 60% s&p. tldr; verticals for startup wins, horizontals for giants. bet moats: data, switches. seen it in

6/ saas shift: legacy horizontals bolt ai. ai-first vertical saas unlocks specialized value. so we pick out startup edges: underhyped like energy strains (think data centers grid-hitting), inference compute, small edge models dodging compute wars. implications: productivity

6/ vertical ai shines: niche models dominate health, finance, manufacturing. bessemer's 2025 state of ai says verticals upend horizontal records via proprietary loops. 70% firms favor verticals for fits. @perplexity is loving horizontal search but difficult in regulated zones

5/ twist: horizontal ai (@openai-style generals) owns broads—@openai's enterprise share at 25%, down from 50%. they gobble horizontal saas, commoditizing search/automation. @perplexity's at $18b val, gunning for $20b, arr ~$80m, bidding $34b on @chrome. but dipping as firms eye

4/ - agents, - inference compute, - mega models, - edge tinies, - endless memory. think realtime market-reading investment bots or ai outpacing docs on rare ills

3/ potentials: agentic ai leads—gartner hype cycle flags ai agents and ai-ready data as top risers in 2025. @ibm sees agents revamping workflows by year-end, from ops gruntwork to custom health. some @forbes spotlights trends:

2/ high-level view: ai market clocks ~$370b in this year, eyeing $2.4t by 2032. @stanford's 2025 ai index notes genai snagged $34b investments, up ~20% yoy. @mckinsey pegs $13t gdp boost by 2030 if deployment nails it, but 95% enterprise pilots tank on integration woes.

1/ been pondering ai's path forward — i think hype's fading, but we're at a tipping point with real-world shifts reshaping sectors. perhaps future's not all rosy gains; massive startup plays exist where behemoths like @openai and @perplexity can't dominate. thread on upsides,

the future isn't in generic ai tools i imagine. it’s in specialized, vertical solutions that solve high-value problems with clear, measurable roi, as demonstrated by the success of companies like frame ai and causaly in their respective industries?

tend to think investment thesis is maturing also. vcs are now prioritizing companies with strong fundamentals and a clear path to repeatable revenue. governance and compliance with regulations like the eu ai act are no longer burdens but strategic differentiators that de-risk

the narrative is also shifting from "bigger is better" to "smarter is better." $nvda research shows small language models (slms) are more effective and up to 20x cheaper for ai agent tasks, democratizing development and creating new arbitrage opportunities.

strategic gap is for vertical ai imho. these solutions are purpose-built for a specific industry or function. their moats are not model size but proprietary data and seamless workflow integration, which leads to lower customer acquisition costs and sticky revenue streams.

i think horizontal ai's vulnerability is its lack of deep domain expertise. while platforms like openai can scale, they struggle to process unique, industry-specific performance signals. they treat content as a simple output, not a performance asset to be continuously refined.

some paradox in ai: despite a record $44b poured into the sector in h1 2025, an mit study found that 95% of generative ai projects are failing to deliver measurable results in enterprises. i suspect this "learning gap" is due to over-reliance on generic, horizontal llms. 1/n

if you’ve got a one-liner sector + two-line defensibility plan, drop it here. i’d love to read.

tl;dr — foundation models = massive upside and distribution power, but they raise the baseline. if you want to survive and scale, own the vertical workflow, own the data, own the compliance, or deliver outcome-level ROI that a hyperscaler can’t copy cheaply.

a few educated hunches i’m keeping attention on: • compliance-as-a-service for ai (audit logs, model provenance, automated evidence for regulators). • tools that let enterprises run private, small task models on-prem or in secure clouds cost-effectively. • operational ai

regulatory angle i imagine will matter more: audit trails, model cards, and certs will be a procurement checkbox in big deals. packaging compliance as a product feature short-circuits a ton of buyer risk — and it’s underrated as a moat.