Akshay

i'm no model training expert, but curious if its garbage in garbage out problem

credits: github.com/openai/codex/c…

couldn't believe it for a sec, but @OpenAI Codex is facing gremlin issues > Never talk about goblins, gremlins, raccoons, trolls, ogres, pigeons, or other animals or creatures unless it is absolutely and unambiguously relevant to the user's query

@AlexanderKalian ditto. this video was a god send, for techies bound by fake limitations youtube.com/watch?v=43Rhhw…

I first truly fathomed the scale of the AI bubble when I tried Y Combinator's cofounder matching service. As soon as I wrote that I was an AI PhD guy, I got swarmed with Ivy League MBA founder teams looking for someone to "do the AI magic" for their startup ideas. One team of three Harvard MBAs told me they had already secured $2M in VC funding to "solve the unit economics problem of AI". They just needed an AI guy to… you know… actually solve it (one of the toughest challenges in AI research right now). In return, I'd get 20% equity while they handled marketing, pitch decks, and whatever else - after the VCs took their cut. I still wonder to this day who gave three Harvard MBAs with zero AI expertise $2M to start an aspiring frontier AI lab. And this is just one of many stories. The situation is not normal or economically efficient. It is blatantly a speculative tech mania and hype-driven investment bubble.

Did you use @ChatGPTapp to write the original mail @OpenAI ?

When we artificially dialed up the “desperate” vector, rates of cheating jumped way up. When we dialed up the “calm” vector instead, cheating dropped back down. That means the emotion vector is actually driving the cheating behavior.

Two years ago, a good YC demo day target was $150k of annualized revenue. In my group office hour today, the lowest demo day target was $800k, and most companies were aiming for $1-2m.

@supabase I shared this on YC's Bookface. Someone asked: "How can I gift this to my dad?" $80 hardware. $0.37 AI. 433 birds. One answer. Full blog + code + interactive dashboard: akshay326.com/bird-feeder/

@supabase Best frame: March 31, 12:08 PM Five sparrows on the railing at once. Also caught what might be a hawk in one frame — with a small bird sitting right behind it. Predator meets prey, captured by a $22 webcam.

I spent $80 and 37 cents to answer one question: When do birds show up at my home? 1,500 photos. 433 birds. 3 days. Here's what I found:

@garrytan and here i thought naming my company X26 would help getting in @ycombinator

We’re renaming the YC spring batches from X25 and (what was going to be) X26 to P25 and P26 — P for Primavera, which literally means “first spring” in Latin-derived languages. The original X was a cute programmer in-joke, but people kept asking “what does X stand for?”, so we’re switching to something that actually says “spring” while still keeping it to a single letter.

@claudeai 2 lessons: 1. how you setup game rules matter a lot. there are 26 neighbours and many combinations to decide cell birth and death 2. seeds produce reproducibility: Explosive tetrahedron: - 2039007139@35 - 3650244864@35 Growth but eventual separation: - 2676159653@35

a fun experiment with @claudeai code -- rebuilding John Conway's Game of Life in 3D some seeds lead to amazing complexity super fast (bottlenecked by your browser's GPU) github.com/akshay326/life… 1/N

@garrytan no more fireball shots in end of batch parties? 😢

I'm giving up drinking because of Claude Code. I need my brain to be maximally pristine so I can sling 10k LOC a day

“Verification, not data scale, is the core bottleneck…” Quite true in a lot of senses