GPT-6 Astra's first 48 hours bring real-world use from architecture to robotics
OpenAI calls Astra the start of the AGI era, but safety experts worry its reasoning can't be fully read
1300
Tag
harness
Claude Code hooks block rule-skipping with code
Anthropic's fifth Claude Academy lesson shows how to set up hook events that enforce CLAUDE.md rules
160
Cognition adopts Fable 5.1 in Devin, cuts coding task costs 54%
On top of Anthropic's new-model price cut, Cognition's own harness delivers the same performance for 47% less
100
Perplexity Computer Shifts Only Sensitive Tasks to Local Mac Processing
Tasks that start in the cloud automatically switch to Perplexity's own 27B model once they reach personal files
110
Prime Agent technical report shows ARC-AGI-3 score jump from 30% to 95.5%
Prime Intellect says the gain came from changing the execution shell around the model, not the model itself. The code is on GitHub.
140
Perplexity unveils Brain, an agent memory system
Perplexity says weaving sessions, files, and sources into a knowledge wiki lifted accuracy by nearly 9 points while cutting token usage 15%
130
Perplexity's local agent beats Hermes, Pi in benchmarks
Perplexity says its 27B on-device model paired with a custom harness outperformed open-source competitor harnesses on knowledge-work, web search, and document-parsing tests
20
Perplexity brings cloud-free agents to DGX Spark
Pro and Max subscribers now get a portable computer that runs the orchestrator, sub-agents, and harness entirely on local hardware.
40
As Fable Gets Pricier, Developers Split Coding Work Across Models Like GLM 5.2
Since the high-cost coding model Fable arrived, developers are increasingly routing different tasks to different models depending on cost and job type
40
ChatGPT Work: 98% of OpenAI staff use it, under 1% of outside users do
OpenAI's push to bring agents to white-collar work beyond coding is exposing a stark gap between internal and external adoption
20
Munder Difflin releases open-source harness that wraps Claude Code to build 'your clone'
Install one download file on your laptop and your existing CLI agent turns into a clone that acts like you
50
OpenRouter: AI Agent Token Use Jumps 14x, Overtakes Humans
Since February 2026, agent token usage has grown 14-fold, compared to just 2.8x for humans
50