Deep Agents v0.6 brings Delta channels, reducing checkpoint storage by up to 100x for long-running agents, without sacrificing observability or resilience. Here’s a 200-turn coding agent session.
Without Delta Channels: 5.3GB of checkpoint storage
With Delta Channels: 129mb
LLMS
-
Deep Agents v0.6 Introduces Delta Channels for 100x Checkpoint Storage Reduction
By
–
-
Open-Source Avatars Level Up with AVTR-1 Locally
By
–
OPEN-SOURCE AVATARS JUST LEVELED UP@avaturn_me just open-sourced AVTR-1, and the tech is pretty wild.
— Charly Wargnier (@DataChaz) 27 mai 2026
> clone the repo
> download the weights
> then run a conversational avatar locally on your machine at almost zero cost 🔥
repo + weights + demo links below 🧵↓ https://t.co/IbrBKsDFylOPEN-SOURCE AVATARS JUST LEVELED UP @avaturn_me just open-sourced AVTR-1, and the tech is pretty wild. > clone the repo
> download the weights
> then run a conversational avatar locally on your machine at almost zero cost repo + weights + demo links below ↓ -
Codex allocates 2% context, warns when adding skills
By
–
Problem is that codex gives me 2% of my context so random stuff adding skills gives me warning messages.
-
AI Agents Have Three Memory Layers: Working Memory
By
–
You use AI agents every day.
ChatGPT, Claude, Gemini, Cursor. They remember what you said, who you are, and what you've worked on before. But this memory has three layers, not one. Working memory
Your current chat. Whatever you typed five messages ago. Gone when you close -
Top AI News: AGI, Bias, Tools, and Industry Insights
By
–
Top stories in AI today: – Exclusive interview with Demis Hassabis on AGI
– Jensen Huang pushes back on ‘AI-proof’ subjects
– Build a weekly marketing report that runs itself
– Stanford finds clear racial bias in AI hiring
– 4 new AI tools, community workflows, and more -
AI Model Pricing Wars and Sustainable Engineering Behind the Cuts
By
–
DeepSeek just made its 75% price cut on V4-Pro permanent. Xiaomi's MiMo slashed V2.5 pricing by up to 99%, effective today. Most coverage frames this as a price war. The more interesting part is the engineering that makes these numbers sustainable. DeepSeek's V4 paper describes
-
Apple Watch App Records Chats, Connects to Claude for Instant Conversation Database Queries
By
–
PULLING OUT YOUR PHONE TO RECORD A CHAT IS AWKWARD@bluedot_ai just completely solved this with their new Apple Watch app.
— Charly Wargnier (@DataChaz) 27 mai 2026
The magic?
It connects seamlessly to Claude.
So you can instantly query months of convos/meetings like a database 🤯pic.twitter.com/Q0vbf4AH3D https://t.co/Uw7qJ8bGg2PULLING OUT YOUR PHONE TO RECORD A CHAT IS AWKWARD @bluedot_ai just completely solved this with their new Apple Watch app. The magic? It connects seamlessly to Claude. So you can instantly query months of convos/meetings like a database
-
Future of AI agents lies in smart model combinations, not size
By
–
The future of AI agents may not belong to the biggest models. It may belong to the smartest combination of models. A lot of companies are currently using large language models for everything:
summarization,
classification,
tool routing,
structured outputs,
simple repetitive -
Perplexity open-sources Bumblebee (blog + GitHub)
By
–
Blog post + GitHub repo: http://
perplexity.ai/hub/blog/perpl
exity-is-open-sourcing-bumblebee
… http://
github.com/perplexityai/b
umblebee
… -
MiniMax releases M2 series technical report for agentic tasks
By
–
MiniMax just released a technical report for the MiniMax-M2 Series! The MiniMax-M2 series, a family of Mixture-of-Experts models designed from the ground up for agentic tasks. The flagship M2 packs 229.9B total parameters but activates only 9.8B per token—thanks to an