kkm
Re:Aleph Alpha's latest Model playing Doom
on Saturday October 10, 2026 @01:45PM (#50035284)
Thread on how we got it running:
https://x.com/konarkmodi/status/2107577982801834136?s=46
Aleph Alpha's Kolibri-1 playing Breakout – Making decisions under 25ms per move
Re:Kolibri Has Landed: A Sovereign Open-Weight Model
on Saturday October 03, 2026 @02:59PM (#49946777)
I agree.
Re:Kolibri Has Landed: A Sovereign Open-Weight Model
on Saturday October 03, 2026 @02:59PM (#49946772)
Correct, search is added as a tool.
Uses Brave search, the idea is to test how well the Model can decided when to leverage search or not.
Re:Kolibri Has Landed: A Sovereign Open-Weight Model
on Saturday October 03, 2026 @12:34PM (#49945684)
Thank you for the feedback on UI, improved the streaming to make it less frustrating.
Re:Kolibri Has Landed: A Sovereign Open-Weight Model
on Saturday October 03, 2026 @10:12AM (#49944434)
Thank you Aleph Alpha team for making it open.
We as many other’s were curious to try and benchmark it.
On that note, as a small gesture of support, we’ve hosted and made Kolibri-1 free for anyone to try for the next few days.
No GPU. No setup. Just try it. tesseracted.com/kolibri-1-chat/
How to serve trillions of tokens for trillion-parameter coding agents
Unlocking parallel test-time scaling for long-horizon agents
Claude.md is good for taste and project context.It's a weak place for invariants
Moving coding-agent guardrails from prompts to hooks
Improving Throughput by Optimising KV Cache Efficiency for Agentic Workloads
Guide to the Kimi DeltaNet Family of linear attention
Forensic Analysis of Container Snapshot Chains for Post-Event Reconstruction [pdf]
The Agent swarm that designs itself
Don't Build a Router. Train the Small Model to Know When to Defer
The gap between open weights LLMs and closed source LLMs
InfiniBand, RoCE, and All That
2678x Faster Matrix Multiplication with a GPU
UCCL-EP: DeepEP-style expert parallelism on any NIC, no GPU-initiated comms
Hacking Google with A.I. For $500k
How to setup a local coding agent on macOS
Anatomy of a high-performance EP kernel
No Token Left Behind: Demystifying Token-in-Token-Out in Miles
MoE expert co-activations: Reordering inputs yields easy throughput gains
The Economics of Speculative Decoding
Speculative KV coding: losslessly compressing KV cache by up to ~4×
70x faster cold(ish) starts for SGLang
Re:Bringing Up DeepSeek-V4-Flash on AMD MI300X
on Tuesday June 02, 2026 @04:51PM (#48376094)
This is very interesting, planning to write about it?
Re:Bringing Up DeepSeek-V4-Flash on AMD MI300X
on Tuesday June 02, 2026 @03:30PM (#48375057)
Also the vllm patch accompanying the blogpost:
https://github.com/doublewordai/vllm-amd-blog-doubleword