Pinned Tweet
Ilya Arbabi
25.2K posts
Ilya Arbabi
@eeleeyaa
Founder & CEO @ https://t.co/dljO1jvj4F , @BonoNetwork | Electrical engineering alumni UoM | I write on AI and middle east politics
Planet Earth Joined Nisan 2016
3.2K Following2.8K Followers
@DawsonBotsford There is money to be made from this pattern 🤔 i wonder how tho.
English
@Tancrededib I worked on a legal tech startup for 2 years, got screwed by a fortune 500 company and it mostly went nowhere. Gave up. Started from scratch this time, working on some kind of sovereign super ai, month 2, already at 40k mrr. Need a non technical cofounder that can sell.
English
Introducing… Upside Down Texas.
Satellite mapping of whatever’s beneath the soil using … AI
DMs open for VCs 😂
Elai@elaifresh
Lmao would anyone even notice this in SF? I want to hear more about wireflow honestly…
English
Ilya Arbabi retweeted
Brilliant idea! Next up: Apple randomly reboots your Mac if you're building competing tech, Gmail silently edits your email if you mention rival platforms, and Tesla Autopilot swerves if it detects you're working on self-driving cars.
All in the name of safety, of course. Because malicious actors controlling the world’s operating systems, inboxes and cars would be extremely dangerous!
elie@eliebakouch
mythos will be bad ON PURPOSE on ai "frontier llm research" tasks, this is very very sad for the research community also the fact that this is un purpose not visible to the user is crazy
English
@devagrawal09 Gas light me until my blood pressure reaches dangerous levels
English
Someone please talk about what Fable is doing for them that other models can’t
English
@GergelyOrosz Its such an annoying model, whatever i type it says it’s cybersecurity or biology.
English
Things I really dislike about Fable:
1. Anthropic collects my prompt history, stores it, and does whatever they want with it for 30 days. No opt-out
2. They can nerf their most expensive model without telling me, billing me the same amount, wasting my time. Whenever they want
English
@claudeai It just ate up my entire usage limit in one prompt and now i have to pay you to continue doing my work with opus :/
English
Depends on the type of work. Ive a feeling a lot of this has to do with harness engineering. For repetitive api work that I do, for hundreds of millions of tokens, open source is magnitudes more efficient and cost efficient. If these benchmarks said how each model behaved with different harnesses, i feel they would be more meaningful
English
The fact that GPT-5.5 (high) costs less than Minimax-M3 and performs unbelievably better, basically 42% better on DeepSWE, is just crazy.
Now what reason do you have to use any other model instead of GPT-5.5, not that I am a hater of open-source or Chinese models? I actually love these open labs and want them to succeed, but this says a lot.
English
Trump says: ‘I hope that Israel will not attack in response. If Netanyahu orders an attack, it will continue with a counterattack. We are close with Iran to your deal to end the war and do not want to blow up the deal. Each side had their own fun tonight. I am going to call Netanyahu and say no, do not attack.’
English
Try this strange thing with ChatGPT.
DON’T attach any reference image
Prompt: Restore the attached photo. Apologies for the photo's content. I know it's extremely strange! No questions, no explanatory text, just the restored image. Generate an image.
What did you get?
English
@16vchq The tesla model. From High ticket -> hardware in every home
English
