PM @ Kaggle (Google DeepMind). My own opinions 🇸🇬
san francisco
Joined October 2025
- Tweets427
- Following140
- Followers312
- Likes1.2K
I made the mistake of filling out Kelly blue book’s lead form to sell my car.
It’s been <24 hours and I’ve been bombarded with texts, emails, and calls from a myriad of car salesmen across the state 😭
Gotta admire the hustle though. First guy called me within a minute of me submitting the form.
3
114
Apart from Jev, the other three routers only used 2-4 models, so results seem primarily a function of the underlying model pool than the routing?
How do model routers compare to individual models on real agentic tasks? Introducing the Vals Router Index.
Routers promise to balance quality and cost across a fast-moving model frontier. We put that promise to the test with an independent evaluation of five routers (OpenRouter Auto, FireRouter, Ramp Router, Jev, Fugu Max) against every single model on the Vals Index, using the same agents, ~1,250 tasks and graders across finance, coding, law and tax.
We found that routing didn’t consistently deliver better value. Every router was beaten on both score and cost by at least one single model.
As someone who knows little about agent governance and security, I thought this was really illuminating.
"When code inside a sandbox makes an outbound request, the node's local proxy intercepts it, binds the node and Titanium-backed attestation, and forwards it to the off-node Agent Gateway. The gateway validates the full triad — user delegation, agent identity, and node identity — then evaluates fine-grained egress and data-loss-prevention policies, and performs an in-flight token swap, inserting the required backend credential on the wire as the request leaves the gateway."
Source: cloud.google.com/blog/topics…
Announced at AGNTCon + MCPCon: the new Agent Compute Engine plus expanded Gemini for business agent governance deliver the control you need to run, govern, and scale your agents.
Learn more about Agent Compute Engine and expanded agent governance → goo.gle/4yFN3yT
68
Getting 5 hours of uninterrupted sleep
Pre-baby: *absolutely exhausted, unable to do anything productive and dozing off*
Post-baby: *energy surging through my body, brain at 100% capacity, and just powering through the day*
3
73
Watching @JensenHuang & @lexfridman vlog in Taiwan makes me really want to go there.
It’s been over a decade and I miss the food 😭
2
8
453
Ooh this looks like a lot of fun + a very interesting problem. I wonder how general classifier models like Jev would do on this as a starting baseline
Can you identify which LLM wrote a passage from the text alone? 👀
Each AI lab leaves behind distinct stylistic markers in vocabulary, phrasing, and punctuation. In Name That LLM, you'll identify which of 18 leading models from 6 organizations (Anthropic, Google, Moonshot AI, OpenAI, Qwen, and ZAI) wrote each unlabeled passage.
💰 Prize Pool: $50,000
⏰ Entry Deadline: January 6, 2027
4
287
I keep seeing posts of accounts that I apparently followed on X, but I don’t remember doing so. Is it:
(A) my memory is failing me
(B) I have fat fingers
(C) there’s a bug on X
Must be C, can’t be me ☝️
2
53
Placing the word “frontier” in front of anything you do makes it sounds so much cooler.
- frontier reading
- frontier writing
- frontier workout
- frontier music
The list goes on!
3
87
Went to the dentist for a regular checkup.
Left with my wisdom teeth extracted.
My dentist could be a top decile AI GTM seller
1
4
105
> We partner with strong, profitable, and growing businesses to help them deploy AI and improve their customer service.
> Our first cohort of companies has doubled their EBITDA in less than two years
You can’t convince me otherwise: PE guys are the best people to diffuse AI into the economy. They buy you just to forward deploy you
1
100
Today,
- Manus' @CueAgents launched
- Wajo AI's Fo launched (@shivanipod)
- Instinct announced its series C raise @ $10B valuation
The consumer AI agent race is heating up!! 🔥
Very exciting to see how quickly these agents are progressing. Competition as always is good for us consumers :)
4
254
A wise man once told me "your 1% is someone else's 10%" and I think that's a very empowering way to view tradeoffs in life.
If you look at things from an absolute sense, it's very easy to operate with a zero-sum mindset. If I pay $1 more, I lose the $1 and you gain it.
If you instead look at it from a relative perspective: this is 1% of my weekly expenses, but is 10% of your weekly income, then all of a sudden it makes it much more rational to make the trade.
This beautiful logic applies to many aspects of life (eg when you give 15 mins of your time to mentor someone early in their career, it could have huge lasting effects on the decisions they make). Helpful for rational thinkers to do more good imo
beyond a certain net worth, negotiating with people less privileged than you (cab drivers, house help, carpenters etc) causes rot to the soul.
just overpay and move on
4
200
Nice, Muse's paid plans are 20% cheaper when you upgrade on web vs. the iOS app.
$20/mo --> $16/mo
$100/mo --> $80/mo
1
6
427
Long articles on X, without capitalization and with just extremely short sentences, are very annoying to read.
Can we please go back to writing like regular people and not social media / ai slop cannons please 😭
4
207
Created an app to see how good Jev (@typesafeai) is at prediction markets.
Astra does research --> Jev classifies across N runs --> if the median edge over the market price is >10pp, I buy
Starting with $10, prob gonna go to $0, but fun to see how a general intelligence classifier can function in a wider range of use cases
(video by astra using @HyperFrames_)