Fsbonetto

Teaching LLMs to think ahead; Learning about ML since 2017

[Fsbonetto @reasoningTokens]

https://t.co/MoxkNsGZGt

[Fsbonetto @reasoningTokens]

Yup, they will monitor very closely what can and cannot come out of @AnthropicAI and @OpenAI. There will not be a Fable 6 situation.

[Fsbonetto @reasoningTokens]

If you think this is about Fable 5 you are wrong, it’s about Fable 6,7 and 8.
We will never hear about something like this again. The next models will be reserved for a elite, the friends of the king only.

[Fsbonetto @reasoningTokens]

@elonmusk Hopefully will show results in the next grok model. 🙏

[Fsbonetto @reasoningTokens]

@lauriewired Take a look at https://t.co/z6Xyk9UB54 if you are going to fiddle with FPGAs!

[Fsbonetto @reasoningTokens]

@karpathy did a great job with auto researcher. Using a similar harness AI is able to build a @risc_v CPU better than humans using @YosysHQ: https://t.co/x9DHH3O9Rd

[Fsbonetto @reasoningTokens]

@gdb Its great at designing CPUs as well. Pairing with a research loop it overtook the human benchmark in 6 hours: https://t.co/x9DHH3O9Rd

[Fsbonetto @reasoningTokens]

@wayne_m159 How the fuck he took the hands out of the steering wheel

[Fsbonetto @reasoningTokens]

@AnthropicAI tried to slide Sonnet 4.6 on my MAX subscription. Coders beware, its not only usage limits anymore

[Fsbonetto @reasoningTokens]

@pmarca @stratechery @benthompson They just added 10Bi of revenue with opus 4.6 last month. Their compute hasn’t grown at the same pace. Money is easier to raise than factories

[Fsbonetto @reasoningTokens]

I’ve being poking around with agent harness and fully autonomous companies and I think I’ve seen end game

[Fsbonetto @reasoningTokens]

@sama Hey bro, when we will have a response to Mythos? AI should't be gatekeep from humanity. Some real aura farming possibilities here

[Fsbonetto @reasoningTokens]

@pmarca Gatekeeping is the most dangerous thing in AI

[Fsbonetto @reasoningTokens]

This is when AI becomes truly dangerous: Gatekeeping.
The moment that not releasing a model becomes more financially beneficial than releasing it. Between doomerism and marketing, the release or rather lack of, Claude Mythos is the first model that really spooked me. Not because of its capabilities, but because of the gatekeeping. Yes, Anthropic is playing the good guys here, but what if they aren't? What if they gatekeep the next model to Mythos for their own benefit? They would pull so far ahead of society and other software companies, would acquire so much power (through hacking) and money (through arbitrage and trading strategies in the financial markets), that the next model might be the final frontier. The AI 2027 vision doesn't look so distant after all.

[Fsbonetto @reasoningTokens]

@elonmusk Are you guys using Distilled Pretraining https://t.co/zhxNXgwPeQ
Or are you wasting the tokens of the 10T model and not using it for the 1.5T?

[Fsbonetto @reasoningTokens]

@pmarca Price will go down by itself. Better hardware, better models, better inference optimization. How much Opus level intelligence will cost 1 year from now?
Making the systems run reliable will be a greater challenge IMO

[Fsbonetto @reasoningTokens]

CLI tools is quickly becoming the standard for agents. If your company don’t have a cli tool, you will be left behind

[Fsbonetto @reasoningTokens]

@naval There was a time that a chess player + a chess engine would beat the chess engine. That time in history was very narrow