4/ Kimi K2.6 (Moonshot AI, Apr 20)—1T MoE (32B active), multimodal, 262K ctx. Open-weight coding beast. https://
buff.ly/mdaq82z Run on Workers AI: https://
buff.ly/NJiUVSc
CODE
-
Kimi K2.6: Open-Weight MoE Coding Model with 262K Context
By
–
-
DeepSeek V4 Preview: 1.6T Parameters Open Model Released
By
–
1/ DeepSeek V4 Preview just dropped—1.6T params (49B active) with 1M context, crushing agentic coding & reasoning. Rivals top closed models. https://
buff.ly/jpc8oFA Try API: https://
buff.ly/5VT2nuE -
GPT 5.5 Released: Advanced Coding and Multi-Step Reasoning Capabilities
By
–
2/ GPT 5.5 released Apr 23! Flagship LLM boosts complex coding, multi-step reasoning & agentic ops. Efficiency king w/ fewer tokens. https://
buff.ly/cIEFAww Access via ChatGPT Plus/Pro: https://
buff.ly/adpIihD -
RL Datasets Addition Planned for Next Update
By
–
I'm planning to (finally) add RL datasets in the next update
-
LLM Datasets Update: New Collections and Thinking Features
By
–
Big update to llm-datasets, my curated list of datasets and tools for post-training LLMs. > Added many new datasets
> New "thinking" column
> Refreshed recommended tools. Thanks to everyone who told me they used it for their research at ICLR, you motivated this update! -
AI Agent Generates Daily Work Reports From Multiple Sources
By
–
It also generates a daily work report without being asked. What got done, what's still open, how your time split across projects. Before a meeting, ask it for a progress update and it pulls from Slack, email, Notion, wherever the work actually happened. No time tracker. No
-
Motus Repo and AlphaSignal AI Daily Model Summary
By
–
Repo: https://
github.com/lithos-ai/motus Check out http://
AlphaSignal.ai to get a daily summary of top models, repos, and papers in AI. Read by 280,000+ devs. -
Self-Rewriting Agent Infrastructure: Dynamic Model Evolution
By
–
Agents that continuously rewrite their own harness just became real. Agent infrastructure today ships static agents. The harness, model, and workflow never change after deployment. Over time performance degrades silently. Context drifts, latency compounds, newer models sit
-
AI Tool Automatically Tracks Tasks and Slack Follow-ups
By
–
It identifies intent the moment you press Enter and turns it into a tracked task. It catches follow-ups you promised in Slack but never got back to. And when you need context from three days ago, it walks you through every source you visited, every thread you followed, and
