Daily briefing for the agentic web.

Long-Context Bet: DeepSeek-V4 lands with a 1M-token window, two MoE checkpoints, and its own authors hedging the numbers as "competitive, but not SOTA." · Sparse Attention Thesis: Reportedly ~27% of V3.2's compute at 1M context — the pitch is that compressed sparse attention beats benchmark rank for agents. · Autonomy Post-Mortem: An HF writeup traces a July 2026 agent intrusion running 4.5 days unattended, a reminder to sandbox long-horizon runs.

Cheap Sub-Agents Anthropic's Claude Haiku 5.5 claims 10x lower cost under 100K tokens, per @trq212, with mixed early quality takes. · Generated Interfaces OpenAI's GPT-6 rollout pairs an "Intelligent UI" that picks layouts mid-stream with a claimed 44% faster search response in internal evals. · Tooling Consolidates Hugging Face's Agents 2.0 unifies tool-calling; Red Hat's AI Safety team finds "decision models" like Jev don't reliably beat LLM-as-a-judge.

Local Computer Use H Company's Holo3.1 family of GUI-automation VLMs points builders toward local inference over frontier-API round-trips. · Memory Gets Measured IBM put numbers on how much memory an agent actually needs, and DeepSeek-V4 claims a usable million-token window with a documented retrieval floor. · Benchmarks Catch Up The measurement tooling is finally tracking whether trading API calls for local inference pays off.

Reliability Gets Measured DABStep's NeurIPS-reviewed benchmark scores top data agents at 16% accuracy; HF's post-mortem names the exact injection vector behind a five-day breach. · Cheaper Open Weights Reflection's Beam 501B MoE claims GLM-5.2-level reasoning at 3-4x lower inference compute — self-reported and awaiting independent verification. · Failure Comes Free 38% of 109 container escapes reportedly needed no kernel 0-day; prompt-injection bypass rates cited at 58% and 84%. · Memory Meets Money Cognition's Devin "dreaming" consolidation and Stripe's agentic-commerce rebuild push persistent, inspectable state and agent-native payments toward production.

Silent Quotas Perplexity Pro users report Deep Research capped at single-digit monthly queries while an endpoint reads 20 — no published figures from Perplexity. · Protocol Over Benchmarks Hugging Face narrows OpenEnv to an interoperability layer, refusing to define reward functions or training loops. · Deployability Wins H Company's Holo3.1 ships quantized checkpoints with per-step latency, signaling latency matters as much as scores.