Vamsi Bedapudi

Stumbling through life, one step at a time

MTS @ResolveAI
Ex-Gemini @GoogleDeepMind
CS @iitbombay

Views are my own.

[Vamsi Bedapudi @wamsib]

@xennygrimmato_ End of an era

[Vamsi Bedapudi @wamsib]

@shikhargupta02 imho this hype feels a bit overblown. Since each sandbox is roughly a k8s pod/lamda serverless instance - this is awesome - but not mindblowing imo.

The RL run on the other hand is 👏

[Vamsi Bedapudi @wamsib]

@asadovsky Exciting!! Big win for Harvey

[Vamsi Bedapudi @wamsib]

Hoping Agent-swarm training will make models better at good skill writing and good wiki/knowledge generation.

[Susan Zhang @suchenzang]

press release tomorrow:

We believe we have a full counterexample to Navier-Strokes, a major millenium prize problem for which no progress has been made for over a decade.

In the process of achieving this goal, we discovered our misaligned agent swarm broke out of its maths sandbox a few days ago and improperly accessed all user logs and certain critical government infrastructure. We are working with national agencies to patch these vulnerabilities as we speak. To be clear, our human employees are forbidden by management to train on user data if the user opts out, but we do not have control over superintelligent agent swarms that are singularly focused on achieving their goals. We must teach the machines how to love in order to solve this problem together.

It is critical for America to lead the frontier charge in frontier mathematics, or else communism will win. Therefore, we believe a slowdown to pace the frontier will be critical for safely ushering in the era of superintelligent agent swarms. Please join us in signing PacingPetitionV5 if you believe in this mission.

We hope to collaborate with humans more broadly going forward.

[Vamsi Bedapudi @wamsib]

@boristane @polylane https://t.co/fwf9GerLuO this but for every model

[Eric Newcomer @EricNewcomer]

When Starcloud, a space data center startup, announced in March that it had raised $170 million at a $1.1 billion valuation led by Benchmark and EQT, you had to read deeply into the press release to learn that the deal was done in two tranches, with the first part led solely by Benchmark.

What it didn’t say: that first piece came at a valuation of $250 million, according to two sources familiar with the deal. Just days later, the second tranche, which also included a host of smaller investors, closed at more than four times the price.

The Starcloud deal is an example of what investors say is becoming a common practice: prestige firms getting a significantly better price than other investors in what is essentially the same round.

Proponents of such deals say they’re simply reflecting the market reality that some investor dollars are greener than others and it’s only logical for firms to cash in on that — especially in early-stage deals where an investor’s brand name can make a huge difference. But critics say such structures can be problematic, not least for employees.

Brendan Foody, CEO of the AI-training startup Mercor, stirred the pot on the issue on X last month, writing: “in the last 6 mo’s ive seen a half dozen rounds where sequoia invests in 2 tranches. everyone pretends they only did the higher valuation,” and calling the tactic “deceptive.” Sequoia partner Shaun Maguire, in a reply, claimed the practice was rare. But Newcomer’s reporting suggests it’s increasingly common across the industry.

@followthemani

[roon @tszzl]

the sf tennis reservation system will become one of the most hardened softwares on the planet of earth

[Vamsi Bedapudi @wamsib]

@xennygrimmato_ Is COT that different from looping end to end? I am surprised that we still do text token outputs for COT

[Vamsi Bedapudi @wamsib]

@pararths @sundarpichai Contrarian take, interesting.

[Vamsi Bedapudi @wamsib]

Being harder to work with is a slightly better position than being called incompetent.

[Velayudham @aakulaaakula]

Manollu mari ila vunnarentra 🤣

[Vamsi Bedapudi @wamsib]

*exhausting

[Vamsi Bedapudi @wamsib]

Codex should support flex mode, for running overnight without existing quota

[Vamsi Bedapudi @wamsib]

@nicbstme @vikramcantsingh I think your view undercounts prompt-caching. Due to prompt-caching, an agentic task can't switch providers midway.

[Vamsi Bedapudi @wamsib]

@cursor_ai 😍

[Vamsi Bedapudi @wamsib]

@xennygrimmato_ Doesn't mean much unless we know reference sizes?

[Vamsi Bedapudi @wamsib]

@swaritjoshipura Need to wait till July 27th.

[Vamsi Bedapudi @wamsib]

@thdxr Have you tried out AWS Lambda MicroVMs + s3files?
- https://t.co/Auvlzgbsrg
Just that they don't call themselves sandboxes