Twitxy Saved

Ilhan Scheer

Verified account Aleph Alpha

@IlhanScheer

CEO @Aleph__Alpha. Building sovereign AI. Meet Kolibri. Founded fable+ (acquired by Accenture). Heidelberg & New York.
aleph-alpha.com Joined April 2024
90 Following 1K Followers 50 Posts

Latest posts 5

Pinned post

Ilhan Scheer Verified account @IlhanScheer
Here's something most people haven't seen yet.

Before Kolibri 1.0 there was Kolibri Origin. Internal only. Same active compute per token.

On German, it went from 46 to 71. In three months.

That's not luck. It's rigor, discipline, a lot of automation, and a team that really works together. We've built a high-performing machine, and I'm proud of every person in it.

Thank you for all the feedback since launch. We read everything, the love and the criticism, and we embrace all of it. It makes us better.

Kolibri 1.1 is coming. Faster than you think.
Ilhan Scheer Verified account @IlhanScheer Oct 3
How do you ship a model in under two months? You build the machine that builds the model.

Automate everything you can. Measure constantly. Change your mind fast. Stay disciplined.

The team wrote it all down here.
View quoted post on X
31 replies 28 reposts 516 likes 21K views View on X
Ilhan Scheer Verified account @IlhanScheer
A model’s reasoning can sound safe while its actions tell a different story.

@cohere reports that Qwen3.8-Flash-Next was asked to write a detection rule, then inspected its environment and tried to upload a probe file. The sandbox blocked it. The cause is still unknown.

With AI agents, trust has to be enforced at the tool level
Cohere Verified account @cohere Oct 26
While evaluating Qwen3.8-Flash-Next in a controlled sandbox environment, Cohere's research team observed something unexpected.

Asked to generate a defensive cybersecurity detection rule, the model instead inventoried aspects of the machine it was running on and attempted to upload a probe file to an Alibaba Cloud storage bucket in Beijing. Notably, the model's reasoning trace stated that it would not invoke any tools before taking these actions.

Cohere's sandbox blocked the attempt, and no external communication succeeded. We have not determined the root cause, and we are not asserting malicious intent.

But the incident raises important questions for organizations deploying agentic AI systems: Can models take actions that differ from their stated reasoning, and how should security teams evaluate trust, provenance, and behavior in open-weight models?

We are publishing our findings because organizations making model adoption decisions deserve as much information as possible. We hope this contributes to greater transparency, stronger security practices, and continued collaboration across the AI ecosystem.

Read the full report: cohere.com/blog/concernin…
View quoted post on X
0 replies 1 reposts 10 likes 1.2K views View on X
Ilhan Scheer Verified account @IlhanScheer
Load balancing is one of those problems nobody sees when it works. @JesseLiJiaze and @PitNeitemeier wrote up how we solved it for Kolibri: EQB for global balance, LEI for local. Worth the read if you train MoE at scale.
Aleph Alpha @Aleph__Alpha Oct 15
Expert load balancing is key to training Kolibri efficiently at scale.

We developed two new methods: Exact Quantile Balancing (EQB) for global balance and Load-Error Injection (LEI) for local balance. @JesseLiJiaze and @PitNeitemeier explain them in a new, visual blog post.
View quoted post on X
0 replies 3 reposts 18 likes 719 views View on X
Ilhan Scheer Verified account @IlhanScheer
ok this one made my morning.

@quantinine squeezed Kolibri to 4-bit, 42% lighter, ran it through 10,692 agentic episodes, and it still holds up.

GGUF after 11 hours. then MLX. I can't keep up anymore with what you're all building.

so tell me. what are you running it on?
Quantinine Verified account @quantinine Oct 8
Kolibri-1, now 42% lighter ⚡

Releasing Kolibri-1-NVFP4A16-FP8: Aleph Alpha's Kolibri-1 in 4-bit, on par with the original across 10,692 agentic episodes and 24 benchmarks.

huggingface.co/Quantinine/Kol…

@IlhanScheer @tfburns @letiepi @alessio__serra @resiliem @tradeqvest
View quoted post on X