yonge bai

founding eng @ Phonely (YC S24)

[yonge bai @baiyonge]

@Ty_Drums https://t.co/mnRETQiZkI super smash brothers melee

[yonge bai @baiyonge]

Muskman realizing that the personal assistant market is bigger than the model market, so it makes sense to make the personal assistant the best it can be

[Elon Musk @elonmusk]

Important note regarding Grok @Bot:

Going forward, @SpaceX will use the best back end model for any given task, including Claude Opus 5.5, MidJourney, Suno and other leading APIs.

Whatever is most likely to give you the best outcome.

[yonge bai @baiyonge]

Software isn’t art

[hungrybox @LiquidHbox]

Hey all,

I removed the post about Rollback Brawl and Rollback P+.

There were things regarding its development that were heavily misconstrued to me, especially regarding @BrawlbackTeam.

I have spoken to Crunch about this, and I will be distancing myself from the project.

I love Crunch and would take a bullet for him - but I want to do what is best for the community - and Rollback solution with vibe coding does not feel correct in my heart, especially one that does not have the blessing of the people who were working hardest on an actual solution for it - the @BrawlbackTeam.

I’ve read a lot of concerns and replies about the work being built upon other work and I want to get to the bottom of what happened. They are valid.

AI has no place in the world of gaming. I want to make a point to support that moving forward.

I am not a software developer. I did not do my due diligence in fully understanding what was at play here, and the future ramifications it can present. I remember checking if the project was open source because I was curious for a timeline when it would be developed.

I simply wanted to do a favor for my friend for all the things he’s done for me.

I will be reaching out to the @BrawlbackTeam to understand what exactly had happened, and what can be done to rectify things. All I was told from Crunch is that he had their blessing and they asked to be credited in the post.

I apologize to those who were wronged, and to the people who look up to me to be an influence in the community.

[yonge bai @baiyonge]

No one outside of tech cares about their data. Everyone already assumes their data is already being taken. People will trade privacy for convenience every single day.

[Zain Javaid @zainmfj]

Introducing Core, The Personal Al Computer

The most powerful technology humanity has built should belong to you

Order today: https://t.co/tlNxSu5B5C

[yonge bai @baiyonge]

@tj_littlejohn Love to see u on the timeline 🔥🔥🔥

[yonge bai @baiyonge]

Can we collectively gatekeep this area from all the people flying into SF this week 🙏

[yonge bai @baiyonge]

@HumpedBack0 Bro 😭

[yonge bai @baiyonge]

@mattbergland Where y’all play?

[yonge bai @baiyonge]

Gpt 6 feels like the first model trained to be worse at instruction following. It seems more opinionated about the spirit of the request and sometimes it’s opinion is very wrong. But probably my fav model for more open ended tasks

[yonge bai @baiyonge]

@WillManidis They probably have at least 90% margin on tokens making it far more reasonable

[yonge bai @baiyonge]

- the social network
- moneyball
- the wolf of wall street

The modern male psyche. You’re welcome.

[yonge bai @baiyonge]

@zzzbuggg Lfg, congrats on the launch!

[yonge bai @baiyonge]

What if tokens really are zero marginal cost?

[yonge bai @baiyonge]

NBA games with the 2K angle r gonna go hard

[World Labs @theworldlabs]

Introducing Atlas:

The world's first multimodal world model that generates image and video frames with pixel-perfect camera control and reconstructs them in 3D.

Model the world, move the camera, and simulate space & time.

[yonge bai @baiyonge]

We stopped renting a model for calls a while ago and built one. The part that surprised me was not the speed, it was the quality: 0.779 vs 0.714 (GPT-4.1) on our benchmark. I assumed we would trade something. We did not

I can make that more concise if you want--just let me know

[Phonely @phonely_ai]

Alma is live. An LLM built for phone calls, trained on 10M+ real conversations.

182 ms to first token (GPT-4.1: 490 ms).
206 ms at p99 (GPT-4.1: 2,018 ms).
$0.55 per blended 1M tokens (GPT-4.1: $3.50).

It already runs 100% of Phonely's calls. Benchmark below, losses included.

[yonge bai @baiyonge]

Well shit

[Z.ai @Zai_org]

Introducing GLM-5.3-Flash

- Leading capabilities at a highly competitive price
- Natively multimodal with a 1M-token context window
- A 320B-A18B model released under the MIT License
- Previously previewed as Ox Alpha, running entirely on Chinese AI chips

Blog: https://t.co/tzOmB7gdZP

Available now across all official platforms:

Weights: https://t.co/9LRMahY9Wa
API: https://t.co/VcaQnzYmS9
Coding Plan: https://t.co/Nk8Y98HNhU
ZCode: https://t.co/Peepqv4XSx
Chat: https://t.co/WCqWT0qCQb
AutoClaw: https://t.co/aGEG5HqTTb

[yonge bai @baiyonge]

Nba defenses are so much more complex now. 1980s basketball is as different to 1950s basketball as it is to 2020s

[yonge bai @baiyonge]

Jokic would average 30 assists a game if the nba still had the illegal defense rule, these are good passes but mostly trivial reads

[LegendOfWinning @LegendOfWinning]

If anybody wants to know why Magic is still in GOAT offensive player convos and why many people still call him the GOAT PG... please watch this series

[yonge bai @baiyonge]

There is 0 chance that this model is chinese

[MTS @MTSlive]

SITUATION BREWING: A highly capable stealth model named Ox Alpha has appeared on OpenRouter with a 1 million-token context window and text, image, and video input. It is free for a week. No lab has claimed it. No one yet knows its true identity.

[yonge bai @baiyonge]

The french are so romanic