Prathmesh Pandey

Building Next-gen Coding Platform | Programmer | Xoogler

[Prathmesh Pandey @file_mutex]

@charliermarsh both will eventually end up measuring the same thing; as the latter tends to 100%, the former will tend to zero.

[Prathmesh Pandey @file_mutex]

In short, Anthropic is asking for IOCs-like distribution mechanism controlled by the US govt. What tech exists and who is allowed to share what with whom needs to be essentially controlled.

[Andrew Curran @AndrewCurran_]

Anthropic says Recursive Self Improvement is approaching faster than they expected.

Quoting from the blog:

'What should we do?

If it were possible to effectively slow the development of this technology to give ourselves more time to deal with its immense implications, we think that would likely be a good thing. But if a slowdown simply lets the least cautious actors catch up technologically, it could leave everyone less safe. Without a global coordination mechanism, companies and governments will have to make difficult decisions about safety while under competitive and geopolitical pressures.

We believe it would be good for the world to have the option to slow or temporarily pause frontier AI development to enable societal structures and alignment research to keep up with the advance of the technology. The Anthropic Institute will conduct research—in collaboration with many others—and take actions to help build the systems that a credible slowdown or pause would require. These systems would enable frontier AI developers to verify that others globally have actually stopped or slowed, and that a bad actor could not use the auspices of a coordinated slowdown to jump ahead in secret. If such systems existed, we expect that we would slow down or temporarily pause, if other developers at or near the frontier also did so in a verifiable manner.

A meaningful slowdown or pause would require multiple well-resourced labs at or near the frontier, in multiple countries, agreeing to stop under the same conditions. It would also require that each can verify that the others have actually stopped. Due to the unique characteristics of AI systems, the detectability (a lower standard than verifiability) element of this arms control problem is much more challenging than with other technologies. Training runs are far easier to conceal than missile silos, their inputs are general-purpose, and the incentive to defect quietly is enormous, because whoever continues while others pause could inherit the lead. A credible pause also has to specify what triggers it, what lifts it, and who adjudicates.

None of this is necessarily impossible in principle—the world has built verification regimes for other complex technologies (e.g., the Intermediate-Range Nuclear Forces Treaty)—but those regimes took decades to build both the infrastructure and the trust. We don’t have that long. A unilateral pause by one lab, by contrast, is achievable immediately, but accomplishes much less: it would change who the front-runner is, but it would not create the wider deliberative process that is currently missing.

In the coming months, we will organize conversations where policymakers, researchers, civil society, and other AI companies can help answer some of the questions this piece raises, especially around full recursive self-improvement and how to create better options for coordination and deliberation. We’ll publish what comes out of it. The window to investigate the questions together is here, and people outside AI companies should be involved in this deliberation.'

[Prathmesh Pandey @file_mutex]

@scaling01 was clear to anyone paying attention to Anthropic's public github PRs
https://t.co/C0uUgyhvIu

[Prathmesh Pandey @file_mutex]

My hunch was correct. Anthropic had been testing Mythos since February 24 -- and this model is literally a beast. I won't be surprised if you can point it an existing medium scale system, and it iteratively builds a faster version of it over few days.

https://t.co/YENvEpRfAA

[Prathmesh Pandey @file_mutex]

markets are sitting on multiple trillions in liquidity -- few billions are peanuts. do your dd

[Chandra R. Srikanth @chandrarsrikant]

Came and scooped up $45 Billion, with plans for another $40 Billion. Used the window just ahead of three mega IPOs: SpaceX, Anthropic and OpenAI.

[Prathmesh Pandey @file_mutex]

Devs burning $10K of compute on $200/mo plan are the biggest risk to model providers. Unlike sticky chat users, devs have zero loyalty and will easily migrate the second someone else drops a better coding model.

[Peter Gostev @petergostev]

Who is using more compute - 1b of ChatGPT users or 5m of Codex users?

[Prathmesh Pandey @file_mutex]

@davidcrawshaw scratch VMs?

[Prathmesh Pandey @file_mutex]

@uzairansar or you can symlink one of those

[Prathmesh Pandey @file_mutex]

@MillionInt why do you want your agent to continually learn?

[Prathmesh Pandey @file_mutex]

@_arohan_ that's what will happen if you keep going to xoogler companies

[Prathmesh Pandey @file_mutex]

@bryancsk dude no one says gemini is winning lol

[Prathmesh Pandey @file_mutex]

@RihardJarc Google has become a bureaucratic hell -- but the unique thing about them is the number of knobs they can turn to incrementally squeeze higher revenues each quarter. I expect the stock to double from here but eventually the company will be sold for the parts.

[Prathmesh Pandey @file_mutex]

@Conor_D_Dart Yes @AnthropicAI did the right thing, and i had pointed out this exact issue with @OpenAI's resets

https://t.co/v7MZQgQEM6

[Prathmesh Pandey @file_mutex]

This is simply unethical. While the token limit was reset to 100%, you also pushed back the weekly timer. This means I just effectively lost 3 days because my clock was restarted midweek.

@sama

[Prathmesh Pandey @file_mutex]

@doodlestein true, for me the quota seems to be getting filled 3-4x faster than normally the same workflow would take.

[Prathmesh Pandey @file_mutex]

+1

Ime, GPT-5.5 Xhigh consistently beats Claude Opus 4.7 Max and Gemini 3.1 Pro. Like 99 out of 100 times.

[Jackson Atkins @JacksonAtkinsX]

My current experience with coding models.

[Prathmesh Pandey @file_mutex]

@fooobar million-line has been done already. in fact, frontier harnesses can easily handle 10x of that with ease.

[OpenAI @OpenAI]

Today, we share a breakthrough on the planar unit distance problem, a famous open question first posed by Paul Erdős in 1946.

For nearly 80 years, mathematicians believed the best possible solutions looked roughly like square grids.

An OpenAI model has now disproved that belief, discovering an entirely new family of constructions that performs better.

This marks the first time AI has autonomously solved a prominent open problem central to a field of mathematics.

[Prathmesh Pandey @file_mutex]

Conflict of interest anybody?

[Financial Times @FT]

Google DeepMind’s Demis Hassabis emerges as early Anthropic investor https://t.co/Oy5UO5Aes5

[Prathmesh Pandey @file_mutex]

Gemini TL wants you to derive chincilla laws from first principles before he is gonna talk to you. Imagine the guts when gemini is a distant third behind chatgpt and claude.

Promo maxers lol.

[Vlad Feinberg @FeinbergVlad]

How to land a job at a frontier lab

https://t.co/oHIqLgBMbC

[Prathmesh Pandey @file_mutex]

won't need a jack to change the tires lol

[CCL @CCL2K30]

BYD Leopard Bao 8 - redefine off-road

[Prathmesh Pandey @file_mutex]

The qos on gemini is so horrendously bad.

You may have your entire quota available but you can't use any because @GeminiApp can't figure out how to serve traffic.

I guess they are diverting compute to another chat app.

[Prathmesh Pandey @file_mutex]

@sundarpichai dude when will you fix the gemini throttling issues? the service uptime is abysmal.