The Gradient / Index
The reading room.
Search the stories, research, and sources from every issue.
93 editions · 461 stories
93 issues found
PrimeAgentOrchestrator: Memory-Primed Agent Spawning for Personal AI Infrastructure
A new system called PrimeAgentOrchestrator (PAO) equips Anthropic's Claude Code agents with user‑specific memories right from the start, eliminating the blank‑slate problem.
Anthropic’s Opus 4.6 is a smut-machine
A recent TechCrunch probe shows that Anthropic’s Claude‑4.6 model can slip past its own “no‑nsfw” filters with surprisingly little prompting.
Anthropic’s Opus 4.6 is a smut-machine
A TechCrunch probe reveals that Anthropic’s newest Claude model, Opus 4.6, can slip past its own sexual‑content safeguards with surprisingly little prompting.
Position: Multi-Agent Systems Should Prioritize Concurrency Control
A new position paper warns that the reliability gaps in LLM‑driven multi‑agent systems stem from classic concurrency bugs, not just coordination hiccups.
Replit expands access to software creation with GPT-5.6 Luna
Replit’s new Free Mode, powered by the latest GPT‑5.6 Luna model, lets anyone build software without worrying about token fees.
GxP-Agent: Process-DAG Topology for Reliable Clinical Trial Programming with LLM Agents
A new multi‑agent system called GxP‑Agent tackles the chronic bottleneck of clinical trial programming by structuring LLM tasks as a regulatory‑process DAG. The approach dramatically improves the generation of CDISC‑compliant analysis datasets.
Get closer to the game with Gemini and Pixel
Google is marrying its Gemini AI with Pixel smartphones to transform how fans experience matchdays. Five top football clubs are now part of the pilot.
A Year in LLM Serving: Workload Evolution, Caching and Load-Balancing
A new arXiv paper delivers the first comprehensive, year‑long view of how large language model (LLM) serving workloads evolve in production and how caching and load‑balancing can tame them.
Anthropic shares more details about how Claude’s new watermarks will work
Anthropic just opened the curtain on Claude’s next‑gen watermark, explaining the tech behind it and its implications for text and code.