investor at @greycroftvc focused on AI, data and consumer
San Francisco, CA
Joined November 2021
- Tweets867
- Following1.3K
- Followers509
- Likes4.6K
Pinned Tweet
Sharing some thoughts on an emerging tension between the progress on long horizon agents and the popular business model of selling the work.
tl;dr there's an underdiscussed risk in customer contracts
congrats to the @reflection_ai team!
Introducing Beam: a highly efficient agentic open model with 501B total parameters and 23B active.
- Frontier reasoning efficiency
- Advances the Western open frontier on coding & agentic tasks
- Trained end-to-end from scratch
Full weights release this month.
Learn more about Beam: reflection.ai/beam
there we go!
nitter.cf/noahrshinn/status/2107…
Kishen Patel retweeted
either way discovery isn't the bottleneck
You get a very different view of the role of AI in drug discovery depending on whether you listen to genomics people ("it's a biological world model") versus medicinal chemists ("it's a research process orchestrator").
- srikosuri.substack.com/p/men…
- linkedin.com/posts/ash-jogal…
I’ve noticed this in the most interesting companies I’ve met over the last year. Younger brilliant founders. They don’t see doom in AI’s impact on society, but instead lots of interesting and novel problems and want to bend the world to their vision of the future. Exciting!
The founders behind companies like @sailresearchco, Instinct @noahrshinn and @ConwayResearch (Underdog) that have emerged over the past few months seem to be getting younger — backed by some of the best-known Bay Area funds. Feels like a golden age for builders.
Kishen Patel retweeted
I asked people which neolabs under $10B have the highest talent density:
32 votes: Core Automation (@MillionInt, @_arohan_)
26 votes: Periodic Labs (@LiamFedus, @ekindogus)
20 votes: Flapping Airplanes (@spectorb, @amspector100, @aidanmantine)
18 votes: Standard Intelligence (@G413N, @devanshpandey)
8 votes: Ricursive (@annadgoldie, @Azaliamirh)
5 votes each:
- Ineffable (David Silver)
- Mirendil (@bneyshabur, @HarshMeh1a, @shayan_, @tararezaeikh)
- Recursive (@RichardSocher, @_rockt, @timshi_ai, @josh_tobin_, @CaimingXiong, @jeffclune, @tydsh, Alexey Dosovitskiy)
3 votes each:
- Applied Compute (@ypatil125, @rhythmrg, @lindensli)
- Isara (@ezhang7423, @hegasz)
- Prime Intellect (@vincentweisser, @johannes_hage)
2 votes each:
- Elorian (@AndrewDai, @yinfeiy, @SethInternet)
- Goodfire (@eric_ho, @DanJBalsam, @banburismus_)
- Inherent (@tantumscollins, @edwardfhughes, @LouisKirschAI, @kallyaleksiev)
- Trajectory (@rronak_, @QuantumArjun, @MichaelElabd)
- World Labs (@drfeifei, @jcjohnss, @BenMildenhall, @chlassner)
People couldn't pick a company they work at or founded. Disclosure: I'm a small investor in Applied Compute, Factory, Standard Intelligence, Trajectory and Wafer. I didn't vote.
Continued:
Kishen Patel retweeted
rebranding of zero-shot classification models as decision models wasn't on my 2026 bingo card
but happy people finally discovered them
Kishen Patel retweeted
‼️The Bitter Lesson for context management: Giving LMs unrestricted control over their context beats human-designed SOTA!
Introducing 🩵Context Language Models (CLMs)🩵
- Natively manage their own context
- Treat context as a file
- Learn policies in CLM weights, no harness
So you want to release a frontier model in AD 2026? Here’s a guide:
1) Benchmaxx the hell out of it
2) Announce said benchmaxxed model to the world
3) Cower before the AI god you have just created
4) Say it cannot be trusted to the masses for whom you presumably built said model
5) Psyop users into believing the model must therefore be something they should use
Kishen Patel retweeted
Looks like we have entered the Tupac vs. Biggie stage of VC and AI investing judging by the posts by my good friend @shaunmmaguire, @vkhosla and others that have crossed my feed.
Free tip for those who need to "get in the mood" to write a vitriolic post. Before I mellowed, I used to play this song back in the days before writing my confrontational activist letters. @bhorowitz will appreciate.
youtube.com/watch?v=41qC3w3U…
Trigger warning: NSFW!
Kishen Patel retweeted
With so many companies focused on selling outcomes (vs tokens) the next observation is optimizing the cost of delivery. @quichenpatel explores this as part of an emerging investment thesis.
Kishen Patel retweeted
When frontier labs talk of “aligning” models, ask: to what? If they say values, ask: whose? If it’s theirs rather than the user’s: why?
Kishen Patel retweeted
Indians face far more discrimination than other groups at elite American universities.
Even outside of Instinct and Muse, it definitely feels like a wave of consumer AI companies is coming.
Model capabilities are sufficient at reasonable cost + founders and investors are excited about taking bigger swings in consumer. Lots good stuff to come!
Kishen Patel retweeted
I present to you unintentionally hilarious VC cope. “Win the text message layer” pleaaaseeeee. If there is any winning to be had in text alone it will be won by meta who literally owns WhatsApp and can give it away for free forever. Cost of capital still matters for true consumer business models so much magical thinking going on
People don’t seem to understand that if you win the text message interface layer, you get the easiest access to embed all of their other contacts. Then you have a bot that learns about group preferences. A new Facebook? Hey - you should date this person, there’s a 98% chance she’s your soulmate!! Let me know if you’d like an intro. I’ll just take a % of your future household income