Yash Patil

@ypatil125

Co-Founder, CEO @appliedcompute 🚂 prev: @OpenAI, @Stanford

following
followers

Reflection AI’s Beam 1.0 is out, with AC2 inference and post-training support

Applied Compute says 40 steps of AC2 post-training raised Beam’s score on a long-horizon coding task from 0.07 to 0.215.

1 Source, 4d ago, first seen 4d ago

Muse Spark Open weights released October 15, 2026

finkd says the model runs well on new Mac Studios, crediting work with Apple and John Ternus.

241 Sources, 10h ago, updated 10h ago

Spork post-training recipe is claimed to raise its maker’s chess-puzzle solve rate by 100%

The @appliedcompute announcement calls Spork an efficient post-training recipe and reports a 100% rise in its chess-puzzle solve rate.

7 Sources, 6d ago, updated 6d ago

AC2 adds custom decision-model training

Applied Compute says teams can use the feature for request routing, content classification and anomaly detection.

6 Sources, 10d ago, updated 10d ago

Automatic hint optimization proposed as a step toward continual learning

Applied Compute says it optimizes teacher-model hints using efficient proxies for their eventual training value.

5 Sources, 12d ago, updated 12d ago

Have AI models reached commodity status?

An essay argues that falling prices and surging token use favor platforms that control distribution over benchmark-leading labs.

57 Sources, 2h ago, updated 2h ago

Training platform aims to automate research work and speed up training iteration

A builder says observing training and inference rollouts helps show how models behave and which behaviors are undesirable.

3 Sources, 18d ago, updated 18d ago

Harvey seeks a lead to help law firms build their own models

A hiring announcement for Harvey’s Private Model Program says the company has built its own model and now wants to help law firms do the same using private data.

2 Sources, 26d ago, updated 26d ago

Startups reportedly turn to open-weight AI to cut costs and reliance on OpenAI and Anthropic

Polymarket, citing Bloomberg, says startups are increasingly using open-weight alternatives to build custom AI models, aiming to spend less and depend less on OpenAI and Anthropic.

2 Sources, 29d ago, first seen 29d ago

The AI software stack beyond model serving

One post argues that AI infrastructure follows cloud computing’s playbook: building valuable software around compute. It sees room for multiple providers, not a winner-take-all market.

1 Source, 30d ago, first seen 30d ago