Davis Keene

The $2T telecom industry was built for humans. @AgentPhoneHQ is rebuilding it for agents. One API gives every AI agent its own phone number and a trusted identity to reach the real world. Congrats on the launch, @themeetmodi & @manav2modi! ycombinator.com/launches/QNE-a…

@ycombinator @AgentPhoneHQ @themeetmodi @manav2modi Congrats!

@snwy_me i really miss seeing dalle2; this is so cool

before DALLE-2 was shut down this month, i generated around ~50,000 images from a diverse prompt corpus and did a full fine-tune of SD1.5 on it. the result feels nearly identical to DALLE-2 (left: base SD1.5, right: fine-tuned)

@kr0der Not enough people are prepared for the subsidy regime to end

hypothetical scenario: all the $20, $100, $200 AI coding subscriptions are gone, no more heavy subsidisations what setup are you using now that you have to pay API prices?

@sugarjammi Irving Farm Coffee on the Upper West Side is one of many sanctuaries in the city for remote work. Other places worth mentioning are The Bean in EV, Devocion in Flatiron, Bar Andiamo UES, Variety in Chelsea, Saltwater by Washington Square Park, and Now or Never in SoHo

where to cowork in nyc (too broke for memberships rn) im talking big cafes + spaces

@theo Excellent essay + video! Wrote about some very similar observations over the weekend: daviskeene.com/blog/the-price… Increasingly looking forward to a world where fewer developers feel like their abilities are starting to atrophy because of coding agents.

Coding with agents is a trap, and we all fell for it.

The Delta of Intelligence is Human Data! @cgnorthcutt is Director of AI Research at Handshake AI. In this episode, Curtis shares his unique perspectives on human data, RL envs, four quadrants of the AI market, journey of Handshake and much more. 0:00:00 - INTRO 0:02:08 - The Snowman Effect: From Rural Kentucky to 7 Years at MIT 0:09:50 - Confident Learning, Two Systems of Intelligence, Classical ML Has to Prevail 0:20:37 - Origin of CleanLab and Handshake acquisition 0:28:00 - How does GPT-5 become GPT-6? 0:31:20 - The Fastest growing Data Lab: What makes Handshake different? 0:35:32 - Difference between Good Data and Bad Data 0:38:17 - New kind of Data, IA and AI 0:42:04 - Scaling Coding Benchmarks and Efficiency in Long Horizon Tasks 0:49:12 - On Pre-Training, Misconception around Human Data 0:57:13 - Open Question: Taste and Personality, Quick Fire 01:05:15 - Advice to 20yo : Skill and Obsession

happy mother’s day

deep research in chatgpt is insanely broken; in the sense that it's really powerful and useful, but also, the product itself is just broken (i cannot save final observations to pdf/latex isn't parsing correctly/certain references just refuse to render)

@manav2modi @ycombinator 🫶

mom! i made it to @ycombinator !!

@andruyeung Chicken Insider on the Upper East Side. Thank me later!

I'm on a quest to find the best fried chicken in NYC. So far: 1. Jollibee 2. Pecking House 3. Double Chicken Please 4. Moono But one just topped the list. It's called bb.q Chicken. Yes, it's a chain. If I'm missing anything from this list, please tell me immediately.

My first foray into AI Safety: DOWSING (Detecting OOD and Weird States in NanoGPT) github.com/daviskeene/dow… I wanted to explore white-box monitoring on an architecture that was small enough to wrap my head around. You can check out the repo or run it as a lab on your own.

@tautologer This is more akin to asking a human, "Smile. Now, how many muscles did you use to do that?" You can "feel" all the muscles in your cheeks and face being used, but it's frustrating to realize that something you have complete control over isn't something you can express.

asking an llm how many letters are in a word is like asking a person what wavelength of light a color is

One of my favorites from Scott Alexander: slatestarcodex.com/2015/04/21/uni…

@quasimatt let's grab a drink at the intersection of technology, media, and art later

Technology, media, art… Just a few things I’m interested in.

@ishaansehgal GGs

still processing this tbh

@ian_dot_so code review is all we do now

Is code review dying/dead? Fill me in on what you’re seeing below

@aaaronson a wise man said this once

@dakeenekid NYC is like Claude Code but for AI

nyc is so great bc it’s the only city where u can talk about ai

decided to do my UX design today on a little easel