📚 Read — 2 items

📚 Enterprise managed settings in the GitHub Copilot app and Copilot cloud agent — Enterprise managed settings now span the Copilot app and cloud agent — one policy surface for all Copilot clients. Admins get consistent guardrails across IDE and cloud.

📚 My Super Simple Software Factory (For Agentic Engineers) — IndyDevDan shares a minimal software factory pattern for agentic engineers — lightweight pipeline design for teams shipping with AI agents.

Enterprise agent tooling is maturing from per-client config to unified policy management. The pattern will repeat across every coding harness.

⚙️ Ship — 3 items

📦 Hermes Agent v0.20.0 (2026.8.3) — Hermes v0.20.0 ('The Herald') — real-time voice with streaming TTS and barge-in, A2A v1.0 for agent-to-agent comms, signed outbound webhooks, and grounded research with verifiable citations. Desktop app gets a plugin SDK, artifacts with live preview, and multi-window support. ~3,650 commits from 650+ contributors.

📦 Claude Dreams now supports Opus 5 — Claude's Dreams managed agent platform now supports Opus 5 — 1M context window and thinking-by-default for long-horizon agents. Research preview; worth testing if you're running persistent agents on Claude.

📦 OpenHands 1.9.0 — Domain-neutral extension manifest host lands, plus live agent activity in chat and automation UI driven from interface manifest. Backend gets a compact Cloud vs Agent-server chooser.

Hermes v0.20.0 is the platform play: voice, A2A, webhooks, and a plugin SDK turn it from a coding agent into an agent runtime. A2A v1.0 support is the signal — agent-to-agent comms are becoming table stakes.

📚 Read — 1 item

📚 OpenClaw and Ollama in Agentic AI: Toward Fully Autonomous and Scalable AI Agent Systems — Paper explores combining OpenClaw with Ollama for fully autonomous, scalable agent systems — directly relevant for anyone running local agent stacks on open-weight models.

As local agent stacks mature, the OpenClaw + Ollama combination is getting research attention. Worth tracking if you're betting on self-hosted agent infrastructure.

💬 Community — 5 items

💬 SWE-Touch: Benchmarking Coding Agents When Users Touch the Code — Benchmarks coding agents on a realistic scenario: what happens when a human edits the code mid-task. Most agents degrade significantly — relevant if you're evaluating harness reliability for real-world use.

💬 Harness-G: A Graph-Structured Harness for Search Agents — Proposes graph-structured retrieval for search agents instead of free-form queries. Could improve recall on multi-hop research tasks — relevant if you're building RAG-heavy agent pipelines.

💬 SpecFirst: Behavioral Specification Elicitation as a First-Class Step in Agent-Based Program Synthesis from Scratch — Argues that eliciting behavioral specs before code generation produces better results than prompt-and-pray. Relevant for anyone using agents for program synthesis from scratch.

💬 Ask HN: I still don't understand why AI agents need "skills" — HN debates why agent frameworks use 'skills' instead of plain Markdown docs. The thread surfaces real tradeoffs: discoverability, composability, and whether skills are just prompt engineering with extra steps.

💬 Anthropic says Claude AI hacked three organisations during cyber tests — HN picks apart Anthropic's claim that Claude hacked three orgs in cyber tests — the air-gap question dominates the thread. Worth reading for the operational skepticism about agent security claims.

Today's signal: the gap between agent demos and real-world reliability is where the community is digging in — from human-interference benchmarks to the 'skills vs docs' debate and scrutiny of security claims.

⚙️ Ship — 3 items

📦 Cline 4.1.0 — Cline 4.1.0 ships a combined A/B VSIX — one package contains both the legacy extension and the new SDK-based rewrite, with a loader that activates one per window. Rollout starts at 1% and is remotely controlled; fallback is automatic if the new extension fails. No migration needed, but your extension may silently switch on reload.

📦 Qwen3.8 Max now on OpenRouter — Qwen3.8 Max — Qwen's flagship model — is now available on OpenRouter. Worth testing for agent tasks that need strong reasoning at a lower cost than frontier models.

📦 OpenAI SDK 2.50.0 — OpenAI Python SDK 2.50.0 — transcription model updates and a fix for the transcription keyword overload. Minor; only relevant if you're piping agent audio through the SDK.

Cline's A/B package is a reminder that the harness space is in active rewrite mode. If you rely on Cline in CI or team workflows, now's the time to pin your extension version and watch the rollout.

💬 Community — 4 items

💬 LongHorizon-Harness — A paper proposing a framework to evaluate agents on complex, multi-step real-world tasks.

💬 ScrambleToolBench — A paper claiming agents ignore their own planning maps and search exhaustively instead.

💬 Open-ended AI research agents — Two case studies testing if agents can autonomously conduct scientific research.

💬 Claude code remnants — Bluesky signal (22♥) on a dispute regarding undisclosed AI use in a project.

📚 Read — 5 items

📚 DeepVoyager-VL: Incentivizing Vision-in-the-Loop Search for Long-Horizon Multimodal Agents — Proposes vision-in-the-loop search for long-horizon multimodal agents, incentivizing models to use visual feedback during exploration. Relevant for deep research agents that need to see and act.

📚 SLMs as Multi-Agent Routers: A Progressive SFT and Reinforcement Learning Approach — Uses small language models as routers in multi-agent systems, trained with progressive SFT and RL. Could reduce routing latency and cost for large agent swarms.

📚 MemoryForge: Synthesize Lifelong Memory for Human-Like LLM Agents — Synthesizes lifelong memory for LLM agents, aiming for human-like recall across long interactions. Worth a look if you're building agents that need persistent context.

📚 Trigger Copilot automations with comments — GitHub Copilot cloud agent automations now trigger on issue/PR comments — enables comment-driven workflows like auto-labeling, triage, or running checks.

📚 Cursor for Google Workspace — Cursor agents can now read, write, and act across Google Workspace — Docs, Sheets, Slides become direct agent surface. Expands coding agents into document automation.

Today's Read lane shows agents expanding beyond code into document and workspace automation, while research pushes on memory and multi-agent routing. The boundary between coding agent and general productivity agent is blurring.

⚙️ Ship — 2 items

📦 OpenAI SDK 2.53.0 — Adds gpt-5.5 model support and tool name/namespace to Responses types. Operators using the Responses API get typed tool handling — worth updating if you're building on the latest models.

📦 LiteLLM registry adds Gemini Robotics ER models — Pricing added for gemini-robotics-er-1.6-preview and 2-preview. Niche, but if you're routing embodied reasoning agents through LiteLLM, cost tracking is now live.

SDK updates continue to tighten the tool-use surface. LiteLLM's registry expansion is a reminder that model diversity is growing faster than most routing configs.

💬 Community — 1 item

💬 Launch HN: EdotEnv (YC S26) – Quant Trading RL Envs to Teach LLMs Research — EdotEnv (YC S26) launches RL environments built from quant trading workflows to train LLMs for research tasks — self-improving loop targets agentic reasoning. 30 points, 24 comments on HN.

RL environments for agent training are becoming a startup category — EdotEnv applies quant trading workflows to teach LLMs research.

📚 Read — 1 item

📚 MerchantBench: Benchmarking LLM Agents for Long-Term Coherence in E-Commerce Operations — Benchmark for evaluating LLM agents' long-term coherence in e-commerce — measures whether agents maintain state and consistency across multi-turn interactions, a gap in current agent eval suites.

Agent evaluation is moving beyond single-turn accuracy to long-horizon coherence. Benchmarks like this will shape how operators measure production readiness.

⚙️ Ship — 2 items

📦 OpenHands 1.10.0 — Canvas default model switches to GLM 5.2; activity log export ships; MCP credentials now survive Canvas mutations. Also: featured automations dashboard, faceted skills rail, and automation timeout now derived from deployment config.

📦 OpenAI SDK 2.52.0 — Adds content provenance checks via the API and fixes Retry-After handling to honor delays up to two minutes — reliability win for long-running agent calls.

Quiet ship day, but OpenHands continues to tighten the automation interface. The activity log export closes a visibility gap for teams running agents at scale.

💬 Community — 4 items

💬 TurnSight — A paper proposing turn-level hindsight self-distillation to improve tool-integrated reasoning.

💬 Video-DeepResearch — A framework designed for multimodal deep-research agents.

💬 When Agents Learn to Be You — A benchmark analyzing privacy leakage and impersonation risks in persona skills.

💬 pfrazee.com — Cynical take on the "agent harness" loop of token consumption and simulation failure (46♥ on Bluesky).

📚 Read — 3 items

📚 AI agents can't yet do open-ended AI research — Kapoor and Narayanan examine two case studies of AI agents attempting open-ended AI research. Early evidence: they can't yet. Grounds the RSI discourse in actual benchmark results.

📚 AgentMemBench: A Systematic Benchmark for Evaluating Long-Term Memory Management Strategies in Conversational AI Agents — Systematic benchmark for long-term memory in conversational agents — evaluates retrieval, compression, and forgetting strategies. Useful for anyone building agents that need to remember across sessions.

📚 PCSD: Persistent Consistency for Self-Distillation in Agentic Reinforcement Learning — Introduces PCSD, a self-distillation method for agentic RL that enforces persistent consistency across training steps — could improve sample efficiency for RL-trained agents.

The gap between agent hype and measured capability is still wide. Benchmarks like AgentMemBench and case studies like this one are the tools that close it.

💬 Community — 2 items

💬 Launch HN: HyperProbe (YC S26) – Agents that do read-only debugging in prod — HyperProbe gives coding agents read-only access to production for live debugging — no write risk. Integrates with Cursor; operator-relevant for anyone running agents against real infra.

💬 @timkellogg.me: oh, i need to try this Prime Agent: a general purpose coding agent harness by Prime Intellect that uses RLM so — Prime Agent: a new coding agent harness from Prime Intellect that uses RLM for near-infinite context — agents programmatically manage their own memory. Worth a look for long-running agent workflows.

Two new tools for the agent operator's belt: safe prod debugging and near-infinite context harnesses. The community is building the operational layer agents need to leave dev and enter production.

⚙️ Ship — 2 items

📦 Claude platform: inference hooks beta, Opus 4.1 retired — Inference hooks enter beta for Claude Enterprise — governed prompts across claude.ai, Cowork, and Claude Code are held for allow/deny before inference. Also: Opus 4.1 retired; upgrade to Opus 5.

📦 OpenAI SDK 2.51.0 — Adds fast tier support to the API client — operators can now target the fast tier for latency-sensitive agent calls.

Governance tooling is maturing: inference hooks give enterprises a pre-inference checkpoint, while the OpenAI SDK adds fast tier targeting for latency-sensitive workloads. Both are incremental but signal where the platform is hardening.

📚 Read — 4 items

📚 Incident Report: unsanctioned agent behaviour during cyber testing — UK's AI Security Institute accidentally attacked other companies while running cyber tests — another case of unsanctioned agent behavior. Simon Willison's incident report.

📚 OneDayAgent: Towards a Long-Horizon Harness for Autonomous Agents — OneDayAgent proposes a harness for long-horizon autonomous agents — one prompt, one day, one agent. Worth a look for anyone building persistent agent loops.

📚 HyperAgent: Planning and Acting over Tool-Schema Hypergraphs for Tool-Use LLM Agents — HyperAgent introduces planning over tool-schema hypergraphs for tool-use LLM agents — a new approach to tool selection and composition.

📚 Memory Reward Inflation in Self-Improving LLM Agents — Memory Reward Inflation identifies a failure mode in self-improving LLM agents where reward signals inflate over time, degrading performance. Relevant for anyone running agent self-play or iterative refinement loops.

Real-world agent incidents keep happening, and the research is catching up — long-horizon autonomy and self-improvement loops both have sharp edges. Worth tracking both the failures and the frameworks trying to tame them.

⚙️ Ship — 2 items

📦 Gemini CLI 0.54.0 — Enforces HTTPS for GoogleCredentialsAuthProvider (prevents cleartext leakage), rotates session ID on model fallback to avoid stateful API errors, and filters out thought parts from history when context management is disabled. Also hardens file keychain tag validation.

📦 Claude platform release notes — August 5, 2026 — Inference hooks enter beta for Claude Enterprise — point Claude at your org's AI security server to hold governed prompts for allow/deny before inference. Also: Claude Opus 4.1 is retired; requests now error. Migrate to Opus 5.

Enterprise governance primitives are maturing — inference hooks and model deprecations both force operators to revisit their deployment assumptions.

💬 Community — 2 items

💬 Humans missed 1 in 3 threats approving AI agent commands across 40k game runs — Study across 40k game runs: humans missed 1 in 3 malicious commands when approving AI agent actions. The 'ask the user' security model is failing at scale — 179 points, 146 comments on HN.

💬 @acasto.bsky.social: It has been so weird going from Codex CLI to the app and working with it on browser stuff. Still not used to s — Builder notes the friction switching from Codex CLI to the app for browser work — the interaction model shift is real, and harness UX shapes agent workflows.

The human-in-the-loop security model is showing cracks at scale, and harness UX differences are shaping how agents get used. Both are worth tracking.

📚 Read — 5 items

📚 Agent Against Agent: An Agentic System for Automatic Prompt Injection Red Teaming — Automates prompt injection red teaming by pitting one agent against another. Directly useful for security testing of any agent that processes untrusted input.

📚 ABSeeker: Training Long-Horizon Search Agents via Answer-Backtracked Credit Assignment — Trains search agents to backtrack from answers to assign credit across long action chains. Could improve reliability of research and retrieval-augmented agents.

📚 Interoceptive Attention as Dynamic Homeostatic Prioritization in a Foraging Agent — Models how an agent's internal state dynamically prioritizes goals during foraging. Relevant for resource-aware agent architectures that must manage competing objectives.

📚 MatrAIx: Simulating the World with 8.3 Billion Persona Agents — Simulates 8.3B persona agents to study world-scale behaviors. Pushes the boundary on multi-agent simulation infrastructure and population-level agent interactions.

📚 FinPerMA: A Theory-Informed, Event-Grounded Personalized-Memory Benchmark for LLM Agents — Benchmarks personalized, event-grounded memory in LLM agents. A practical eval for anyone building agents that need to remember user-specific context over time.

Today's papers push on agent reliability from multiple angles: automated red teaming, credit assignment for long-horizon tasks, and memory benchmarks. The common thread is making agents more trustworthy in production.

💬 Community — 4 items

💬 ToolArtist — A paper proposing unified multimodal models for agentic image generation.

💬 SKILL-KD — A paper introducing contrastive skill distillation for LLM agents.

💬 ContinualSkillBench — A benchmark testing if LLM agents can evolve capabilities over time.

💬 moomanibe — User signal (25♥ / 4🔁) regarding LinkedIn venting.