Manu Otel

Senior Data Scientist @ PublicisGroupe

[Chubby♨️ @kimmonismus]

„Europe wants to become the first AI Continent.“

They can't be serious, can they? This has to be a joke, right?

Completely out of touch with reality.

Absurd.

[Super Dario @inductionheads]

Definitely OpenAI’s worst hire

[Manu Otel @manuotel]

Very bad response…

Use the API properly, otherwise your bench gives bad scores, and makes you untrustworthy.

Deleting thoughts is literally bad engineering, now feels hard to take your benchmarks seriously.

[ARC Prize @arcprize]

OpenAI’s internal testing shows that provider-managed conversation state preserves greater continuity across turns and improves performance on long-horizon tasks like ARC-AGI-3. This is a real and useful result. We’re encouraged to see ARC used to identify useful harness design.

ARC’s verified scores use a “no harness” approach to avoid accidental or intentional developer-aware targeting and to fairly compare scores across all providers. All systems receive the same observations, system prompt, and operate under the same action limits. Conversation state is managed client-side using the industry-wide standard interface for LLMs (the OpenAI-style completions API).

We want progress on ARC to reflect true AGI progress, not ARC-specific format training or settings, and we’re actively working with several industry labs, including OpenAI, to figure out how to best incorporate these server-side state management findings into our verified testing setup while remaining fair and consistent across providers.

[Manu Otel @manuotel]

My gym app is vibe coded 🤣🤣

[Manu Otel @manuotel]

Idk why, but this cracks me so hard 🤣🤣🤣

[Justin Joseph @HuNteR_Jjx]

I'm gonna need whatever armor the ladies of the Autumn got on for my Halo Campaign Evolved LASO playthrough 🤣

[Manu Otel @manuotel]

How did this dummy guy disproof some complicated math conjucture is crazy to me.

[Manu Otel @manuotel]

The line “explicit counter example to jc?!?! Impossible. Find it! Find it!” Are basically the experts inside the model having internal dialogue and debates.

Using moe and rl’ed models drives this type of behavior. Always nice to see it.

[deckard @slimer48484]

That prompt works well for producing counter-examples.

Also Fables CoT leaked when I did this. You can read the full CoT below:

[Tony Statovci @tonystatovci]

How people in Europe GENUINELY walk

[Manu Otel @manuotel]

So this is where Newton discovered gravity

[Manu Otel @manuotel]

Finally!! ❤️❤️

[NXT EU @NXT4EU]

Germany has launched one of the world's best open-source AI models.

Soofi S, made by the Soofi consortium, is a 30B parameter model fully trained in Europe and tops the ranking for open-source AI.

Huge moment for Europe, and finally some competition for Chinese open-source AI.

[Manu Otel @manuotel]

Got early access to mythos 5

[Manu Otel @manuotel]

This is my feed in one SS...

We got a bit lazy

[Manu Otel @manuotel]

EU life 😭😭

[Manu Otel @manuotel]

I was today's years old when I learned that left vs right in politics is not gov vs private, is no-hierarchy vs hierarchy.

[Manu Otel @manuotel]

Psychosismaxxing

[𝙈𝘼𝙍𝘾𝙊 @Marco_CMC90]

Dronă românească 🇷🇴 🚀

[TechWithAmol @The_Designnn]

Please tell me I am not the only one

[Jamieson O'Reilly @theonejvo]

"Let me stop here. You're right, and I should have checked with you first. This deserves a proper investigation, not a guess as to why I just nuked 3 continents."

[Abdulkadir | Cybersecurity @cyber_razz]

“A man who has not hit his Claude limit by noon has wasted his morning.”

-Socrates

[Manu Otel @manuotel]

This is ragebait at it's finest.

[Saoud Rizwan @sdrzn]

honestly impressive this 20 line markdown file has 45k stars on github

shows how much alpha the right prompt can have