I just published “How I scaled Machine Learning to a Billion dollars: Strategy” medium.com/p/how-i-scaled-ma…

The good news is that in 5-10 years we will have British Spring

Amazing And the same core idea as 200M funded startup but on commodity hardware Expect more innovation here

Dimes square

You’re being made a fool But why? It’s Amazon, right! Don’t have have investments in Ant?

It’s more than that If they can’t sell the improving model outputs then they are forced to compete (harder) on non ai things Popping the bubble and making them more entrenched. But don’t think for a second Ant didn’t anticipate this

A bit more than observability Also all execution All thinking tokens Building something here

Resisting the urge to: > Start AI for Enterprise Co > Name it Aurelion Labs > Get two LOIs from my Uncle Larry’s old fraternity brothers > Raise $50M seed pre-revenue > Hire an FDE army of recent CS grads who can’t get software engineer jobs > Have them install OpenClaw to automate email-to-JIRA-ticket workflows > Add OpenAI and Anthropic logos to our partner wall because we use their models > Steal an unknown researcher’s paper and release it as a blog from our internal AGI lab > Sign my Uncle’s friends’ enterprises to 3-year ramp-up contracts with 12-month opt-outs and recognize the terminal year as ARR > Vaguepost about our success on LinkedIn until we get 5 more ramp-up contracts > Start having FDEs label their own computer-use data > Bribe a Hayes Valley realtor for Anthropic and OpenAI researchers’ contacts > Start selling them our FDEs’ computer-use data to triple revenue > Raise $100M Series A > Immediately sell secondaries > Beg Sam and Dario to acquire the company to “expand their Forward Deployed Operations” > Get acquired and escape the permanent underclass Alas, I will resist the urge and remain overseeing my 100 agents that are one loop away from discovering AGI

We will have an AI VC when it autonomously funds Fabrice Bellard’s next side project

It’s important to understand that @ShaneLegg likely means rollouts here not RAG

this model is the opposite of mythos. Its small, cost effective, apache 2.0, and locally deployable. This is the way LLMs should go. small, open source, transparent and sovereign vs large, expensive, proprietary and hegemonic

One weird behavior It really likes going upstream and checking other repos Like ~/work/…/another_repo Need to think more seriously about sandboxing

" - Edge case handled: if the duplicate arrives while the original is still mid-pipeline, the linked asset mirrors its status and flips to ready automatically when the original finishes" the thing is, I never asked about edge case, or sampled fingerprint First model I respect

Ha This is lost of most VCs because they are chasing API wrapper not ML people

Also, those who focus on using AI to help improve the skills of themselves and their teams will be diamonds in great demand, since they will be the rare A players in a sea of mediocrity.

the most talented people I ever worked with OBSESSIVELY look at the data. literally #1 skill. steer clear of those who don’t…

RT @suchenzang: if your bread-and-butter consists solely of: - tuning hyperparams/config files - fitting points on a log-log plot - twea…

Absolutely hate entitled engineers who don’t look at the data and think they can just slap a model run on top The data is the work

If you are a @polynoamial follower This should disgust you

I've got an agent in a loop optimizing a renderer with the goal to minimize frame times (and tests to measure). It got times down from 88ms to 2ms and allocations down from ~150K to 500. Sounds good, right? Wrong. This is exactly why agent psychosis is a big fucking problem. As an experiment, I rewrote the Ghostty core render state in Go, with access to identically laid out data structures as Ghostty and the exact same validation tests. I made a purposely naive renderer (simple, correct, but slow). 88ms per frame with 150,000 allocations (horrendous, lol)! I then kickstarted a Ralph loop to bring the frame times down. I told it it can't modify input data structures or the public API or tests (they're correct), but it can do anything else it wants. It got to work. It has worked for about 4 hours. I've spent around $350 on this experiment so far. The results? 88ms => 1.5ms 150K allocs => ~500 allocs Incredible right? Nope. My hand-written renderer I ported has frame times (same benchmark) of ~20us (0.020ms) and 0 allocations in the update path. This is the problem with psychosis and lacking systems understanding. If you don't understand the system, you're going to accept that this is an incredible result. If you understand the system, you'll see better solutions immediately and can do roughly 75x better on throughput. The people who blindly trust agent output are in the former camp. They're sheeple, overdrinking from a fountain of mediocrity. Standard disclaimer: I use AI all the time. I like AI. The point I'm making is to not blindly accept results. Think. Analyze. Learn.