Matthew

building @incandor_ai

[Matthew @matthew_yekell]

@rshanreddy the only fruit I want in my Reels feed!! Congrats on the launch

[Shan Reddy 🍎 @rshanreddy]

America is banning AI in schools.

China is using AI to create geniuses.

Introducing Aristotle: The AI tutor that solves America’s broken education system. https://t.co/hqPJN8cke0

[Matthew @matthew_yekell]

@KanyesThaker

[Matthew @matthew_yekell]

@asterailabs wooooo!

[Aster @asterailabs]

Introducing Aster Inference -- The world's fastest inference API created by AI research agents

We serve the world's fastest inference on GPU:
- OpenAI's gpt-oss-120b @ 644 tps
- https://t.co/vu7snkRF0D's GLM 5.2 @ 281 tps

At Aster, we're automating open-ended research, and we use inference optimization as a task to benchmark our agents against. We're creating a product out of the inference discoveries made from our system.

As our agents discover more, we plan to further improve our inference product and ship new, SOTA AI products.

[Jessica Chudnovsky @jchudnov]

Two papers accepted at the #ICML2026 workshop on the Impact of Memorization on Trustworthy Foundation Models (MemFM), and I'll be presenting both today!

Scale Dependent Data Duplication: (our first debut of this paper), led by @JoshuaK92829 and Noam Levi.

Internal Data Repetition Destroys Language Models: in case you missed it yesterday.

📍Hall A, 3:50-5:00pm (Jul 11)

Hope to see you there!

[Jessica Chudnovsky @jchudnov]

Pass@k and self-consistency work great for math and code; sample more and verify. So we asked: can the same trick scale truthfulness in domains with no verifier? The answer was no.

Excited to share our #ICML2026 conference paper: Truthfulness Does Not Scale Like Reasoning. https://t.co/RCzcFEPAer. I’ll be at ICML in Seoul to present it!

Co-led by @JoshuaK92829 and @yegordb

[Matthew @matthew_yekell]

@diyasabh @jchudnov I too would as well!

[Jessica Chudnovsky @jchudnov]

Flying to #ICML2026 to present Internal Data Repetition Destroys Language Models, an Oral at Foundations of Deep Gen Models Workshop!

Paper: https://t.co/VyLfnivvUT

You might be curious to know what we mean by “destroys”! Pretraining is now data-constrained, and even aggressively deduplicated corpora keep some repetition. We measured what that repetition actually costs in the currency practitioners care about: compute. The answer, in the worst case, is a third of your FLOPs.

[Matthew @matthew_yekell]

@ismaeel_bashir_ Insanely cool!

[Matthew @matthew_yekell]

what if… my agent… and ur agent… 🥺👉👈

[AgentPhone (YC P26) @AgentPhoneHQ]

this is what happens when you give agents a way to touch the real world.

still buzzing from the call my agent hackathon at @ycombinator

[AgentPhone (YC P26) @AgentPhoneHQ]

this is what happens when you give agents a way to touch the real world.

still buzzing from the call my agent hackathon at @ycombinator

[Matthew @matthew_yekell]

@manav2modi @ycombinator @incandor_ai @RosenzweigLuc finally my agent will have a phone!

[Matthew @matthew_yekell]

@ismaeel_bashir_ @ycombinator @incandor_ai @RosenzweigLuc 🐐🐐🐐🐐

[Matthew @matthew_yekell]

@themeetmodi @ycombinator @incandor_ai @RosenzweigLuc our special talent!!

[Matthew @matthew_yekell]

@KanyesThaker @ycombinator @incandor_ai @RosenzweigLuc Couldn’t have been possible without hyper!

[Matthew @matthew_yekell]

@heyhyperai @KanyesThaker @_shalinshah_ my favorite two young lads!!

[Matthew @matthew_yekell]

@ycombinator @AgentPhoneHQ @themeetmodi @manav2modi LETS GOOOOOO

[Matthew @matthew_yekell]

@ycombinator @Mochatradeapp @theblackmanda congrats this is awesome!

[Matthew @matthew_yekell]

@ycombinator @oddpool_alerts @c0delemons @RiteshMalpani Congrats!!!