📚 Read — 4 items

📚 Co-RL: Unsupervised Reasoning Emerges from Diverse Cohort in Multi-agent RL — Cooperative multi-agent RL where peer-derived rewards replace human labels — reasoning emerges without supervised data. Relevant if you're exploring multi-agent training without reward engineering.

📚 SPADE: Self-Play in Adaptive Synthetic Executable Environments — LLM writes its own executable training environments and learns via self-play — environment scaling without hand-built simulators. Points at self-improving agent training loops.

📚 FM-Bench: A Benchmark for Long-Horizon Management with Competing Agents — Benchmarks LLM agents managing a football club over 20 simulated years with competing agents — long-horizon decision-making under competition. Useful eval for agents that need to sustain strategy, not just answer.

📚 Emergence of Agentic AI: A Review on Evolution, Background, Working Principles, Applications, Adoption Factors, and Future Research Directions — Broad survey of agentic AI — evolution, working principles, adoption factors, future research directions. Useful orientation if you're evaluating where agent frameworks are heading.

Multi-agent RL is moving from hand-crafted rewards and simulators to self-generated ones. If you're building agent training pipelines, the bottleneck is shifting from data to environment design.

💬 Community — 2 items

💬 Bounded Agents: Delegation Security for Multi-Agent AI Systems — Agentic Principal Chain enforces session-aware authorization to block harmful action combinations and delegation abuse in multi-agent systems.

💬 Looped Language Models Improve Compositional Tool Calling — Looped LMs use recurrent computation for multi-step tool use, with adaptive inference balancing accuracy and compute.

Security and tool-calling reliability are the two threads the community is pulling on this week — both papers point at concrete failure modes operators will hit.

⚙️ Ship — 4 items

📦 OpenClaw release-publish/bfa7c3d38e6d — OpenClaw release-publish/bfa7c3d38e6d — no release notes surfaced in the feed; check the tag for operator-facing changes.

📦 Codex 0.149.0 — Codex 0.149.0 adds an interactive `codex agents` dashboard for managing tasks, `codex queue` to message existing sessions, and `codex doctor` diagnostics for endpoint/network/desktop issues.

📦 Antigravity Google updated — Antigravity changelog updated — feed excerpt unreadable; check the changelog for operator-facing changes.

📦 Claude platform release notes — August 19, 2026 — Claude API GA's computer use (`computer_toolset_20260801`) with batch actions and zoom, and launches a browser use tool (`browser_toolset_20260801`) that reads the page's accessibility tree instead of screenshots.

Browser automation is consolidating: Claude's new browser tool reads page structure directly, and Codex adds a task dashboard and queue. Operators should re-evaluate screenshot-based browser agents.

📚 Read — 5 items

📚 Position: Multi-Agent Systems Should Prioritize Concurrency Control — Position paper argues multi-agent systems should treat concurrency control as a first-class primitive — worth reading before your next orchestration layer.

📚 EnvHarness: Awakening Static Worlds for Agent Learning — EnvHarness + EnvRigger reshape static environments via programmable plugins to target agent weaknesses — a practical route to harder RL training worlds.

📚 PolicyGuide: From Guarding One Action to Guiding the Whole Workflow for Policy-Compliant LLM Agents — PolicyGuide shifts from guarding individual actions to guiding entire workflows for policy-compliant LLM agents — useful if you're enforcing compliance across multi-step runs.

📚 Self-Evolving Agents as Dynamic Graph Transformation: A Survey and New Perspective — Survey reframes self-evolving agents as dynamic graph transformations — a useful mental model for tracking how agent capabilities drift over time.

📚 FinSkillBench: Evaluating AI Agents and Domain Skills for Investment Management — FinSkillBench evaluates AI agents on domain skills for investment management — a benchmark for anyone building finance-facing agent workflows.

Research day: the throughline is making agents more controllable — policy compliance, concurrency, and harder training environments.

⚙️ Ship — 3 items

📦 Claude platform release notes — August 19, 2026 — Claude API: computer use tool goes GA (`computer_toolset_20260801`) with batch actions and per-member config; new browser use tool (`browser_toolset_20260801`) for driving a browser in a viewport. Both change how agents interact with UIs.

📦 DeepSeek V4-Flash-Vision-Exp: Multimodal API Live — DeepSeek ships V4-Flash-Vision-Exp — multimodal API now live, adding vision to the Flash line. Relevant for agents that need image understanding without the full-size model cost.

📦 Antigravity Google — Faster project switching, redesigned Customizations settings, and bug fixes — Antigravity changelog: faster project switching and redesigned Customizations settings. Minor but operator-facing for anyone using Antigravity daily.

Tool-use surface is expanding fast: Claude adds browser and computer toolsets, DeepSeek adds vision. Agents that can see and click are becoming table stakes.

💬 Community — 3 items

💬 SWE-bench Science: Can Coding Agents Resolve Engineering Tasks in Science? — New benchmark extends SWE-bench to scientific codebases; agents struggle with domain-specific failures and scientific guidance helps only sometimes.

💬 MemTrapBench: Benchmarking Cognitive Traps in LLM Memory Use — Retrieved memories can induce reasoning errors and belief distortions in LLMs; an inference-time strategy helps avoid these traps.

💬 SkillGate: Training In-Policy Skill Selection in Long-Horizon Agents — SkillGate fixes selector credit starvation in agent skill selection by separating outcome credit for execution tokens from local advantage.

Three papers converge on the same theme: agent reliability is still fragile — memory can mislead, skill selection needs better credit assignment, and coding agents stumble on domain-specific tasks.

📚 Read — 3 items

📚 FreeToken — A paper proposing an edge-native MoE serving system to run models like DeepSeek-V4-Flash on consumer hardware.

📚 FIXING Opus 5 — A video critique arguing that prompt engineering remains essential for model optimization.

📚 StreamOPD — A post-training framework using cue-gating and distillation to improve streaming video understanding.

💬 Community — 1 item

💬 Show HN: OzBrain, a shared brain for knowledge between agents and your team — Show HN: OzBrain pitches a shared knowledge layer between agents and humans — agent-first chat as the primary interface, dashboards out. Early, but the thesis is the debate.

The agent-first interface thesis keeps surfacing in community builds; whether shared knowledge layers become the default is the open question.

⚙️ Ship — 4 items

📦 Claude Python SDK v1.0 — August 20, 2026 — Python SDK v1.0 — breaking: HTTP layer moves httpx→httpx2, temperature/top_p/top_k dropped from Messages, legacy Text Completions removed, async `.with_raw_response` now needs `await response.parse()`. Migration guide has before/after snippets for every change.

📦 Antigravity — Remote Control, faster project switching, and more — Remote Control ships — drive Antigravity from another device, plus faster project switching. Google's agentic IDE keeps landing operator-facing moves.

📦 OpenClaw release (tag 1a3aa6a36af7) — New OpenClaw release tagged. Changelog not surfaced in this cycle — check the release page for runtime changes before upgrading.

📦 xAI API release notes — August — xAI release notes updated for August — Grok 4.6 is the current model line. Specific changes not surfaced in this cycle; check the page for model or pricing moves.

Breaking-change day for the Claude stack. If you're on temperature/top_p/top_k params or the legacy completions API, the v1 migration guide is your afternoon — before the next release compounds it.

📚 Read — 5 items

📚 Repo0: Design-Driven Zero-to-All Code Generation — Repo0 generates complete repositories from natural language by evolving a dual-graph architecture as it codes — a step past single-file coding agents.

📚 FACET: Preserving Source Intent and Executable State in Terminal Task Synthesis — FACET builds executable terminal tasks by grounding instructions, solutions, and verifiers in a shared representation that preserves source intent — scalable supervision for training terminal agents.

📚 Hallucination as a Feature, not a Defect: Evaluating a multi-agent architecture to transform speculative language-model outputs into testable scientific hypotheses — Reframes hallucination as a feature: a multi-agent architecture turns speculative LM outputs into testable scientific hypotheses, with evaluation of the pipeline.

📚 Position: Collusion Risks Among AI Reasoning Agents Justify Certification Requirements for Making Market Decisions — Position paper argues collusion risks among AI reasoning agents justify certification requirements before they make market decisions — a governance angle on multi-agent deployment.

📚 Test Agent Changes with LangSmith Preview Builds — LangSmith preview builds let you test agent changes against production traffic before rollout — staging for agent pipelines, not just code.

Coding-agent research is moving from single-file completion to whole-repo architecture generation, while terminal-agent training gets executable supervision. The next harnesses will be judged on repo-level coherence, not line-level accuracy.

⚙️ Ship — 3 items

📦 Devin CLI v3000.5.20 — Devin CLI v3000.5.20 adds `devin rm` session deletion, `devin doctor` subagent profile validation, `/recap` session summaries, and Agent Plugins 1.0.0 compatibility.

📦 Antigravity Google — Faster project switching, redesigned Customizations settings, and bug fixes — Antigravity ships faster project switching and a redesigned Customizations settings page; bug fixes included.

📦 Hermes Agent 0.20.5 — Hermes Agent 0.20.5 rolls up ~323 PRs: Bot Mode group-room threads, foldable summaries, PDF attachments, keyless web tier with 5-vendor rotation, and runtime stall guards from Composio eval findings.

Harness and runtime updates dominate: Devin CLI adds session management and plugin compatibility, Hermes rolls up a large batch with keyless web tier and stall guards. Operators should check their tooling for new session controls and plugin formats.

💬 Community — 2 items

💬 G Elliott Morris: brother's open-source agent harness is model-agnostic, can pit models head-to-head — G Elliott Morris flags his brother's open-source agent harness — model-agnostic, can run models head-to-head. Worth a look for anyone evaluating harnesses.

💬 Ethan Mollick: Codex & Claude Code now handle everyday form-filling — Ethan Mollick: Codex and Claude Code are now good enough at 'fill out the forms I got in email' to save real time — the everyday-usefulness bar has moved past code-only.

Community chatter is converging on harnesses doing non-code work — form-filling, model head-to-heads. The next wave of agent tooling will be judged on everyday time savings, not just repo work.

📚 Read — 4 items

📚 FlashPrefill V2: Block-Sparse Prefill Attention for Long-Context LLM Serving — Block-sparse prefill attention with mean correction cuts long-context prefill cost — relevant if you serve agents on long documents or codebases.

📚 Persona-Guided LLM Agents for Task-Oriented Dialogue — Persona-guided agents for task-oriented dialogue — measures whether persona conditioning improves task completion or just adds noise.

📚 TeachMateGPT: A Multi-Agent Knowledge-Grounded Framework for Pedagogical Assessment Generation from Science Curriculum Materials — Multi-agent framework for generating assessments from science curricula, knowledge-grounded to reduce hallucinated questions.

📚 The Evolution of the Agent Harness — Dan McAteer: models keep absorbing the harness into their weights — the next harness is for human attention, not the model.

Read lane today is about the harness dissolving: models absorb scaffolding, and the remaining interface is human attention. Worth tracking if you build on top of agent frameworks.

💬 Community — 1 item

💬 Security Assessment of DeepSeek Harness with A.I.G: Evaluating Resistance to Indirect Prompt Injection — Indirect prompt injection still lands at notable rates against DeepSeek Harness — taint-tracked, dual-judge evaluation. Worth a read if you run DeepSeek agents.

Prompt injection remains the unsolved problem for agent harnesses. Every new harness ships with the same exposure until proven otherwise.

⚙️ Ship — 3 items

📦 ACP 1.21.0 — ACP 1.21.0 stabilizes terminal authentication and elicitation in the schema; session compaction updates land behind unstable. Operators building ACP clients get stable auth and elicitation primitives.

📦 Anthropic Python SDK 1.0.0 — Anthropic Python SDK hits 1.0.0 — breaking upgrade to httpx2, migration guide in MIGRATION.md. Beta helpers stop warning on output_format. Anyone pinning the SDK should read the migration before bumping.

📦 OpenHands 1.15.0 — OpenHands 1.15.0 adds an LLM provider-connections UI and an automations catalog with script bundles; fixes silent agent profile downgrade and makes the home LLM dropdown override pinned profiles.

Protocol and SDK stabilization day: ACP locks terminal auth and elicitation, Anthropic SDK hits 1.0.0 with a breaking transport change. Treat both as migration checkpoints, not routine bumps.

📚 Read — 3 items

📚 Shared agentic work with GitHub Copilot in Microsoft Teams — Copilot in Teams now turns meeting decisions into ready work and carries context across Copilot surfaces — a shared agentic workflow for teams.

📚 Filenames are the wrong index for Claude Code @ mentions — Sourcegraph's fileSuggestion hook ranks Claude Code @ mentions by symbol, not path — finds the file where a function lives. Practical fix for path-based matching frustration.

📚 JetBrains Context: Repository Intelligence for Coding Agents — JetBrains ships Context, a repository intelligence layer that gives coding agents structured understanding of your codebase — new tool for JetBrains-based agent workflows.

Context is the battleground this cycle — Copilot, Sourcegraph, and JetBrains are all shipping ways to give agents better understanding of your work and codebase.

⚙️ Ship — 2 items

📦 Anthropic SDK 0.124.0 — Files and Skills APIs hit GA; adds computer-use and browser-use toolsets. Operators building desktop or browser agents get stable primitives.

📦 Vercel AI SDK 2.0.0 — Workflow 5 required — Workflow 4 support dropped; install under the beta tag. Also clears partial UI message parts when a WorkflowAgent model-call step retries.

Two SDK moves today: Anthropic stabilizes agent tooling, Vercel forces a Workflow 5 migration. Operators on Vercel AI SDK need to plan the beta upgrade.

💬 Community — 4 items

💬 Taylor Lorenz — Bluesky signal (141♥) on the friction between using Grok for Substack and audience expectations.

💬 PTXBench — A paper proposing a benchmark for LLM performance in architecture-specific GPU kernel optimization.

💬 GenRouter — A routing framework designed to direct prompts to the most efficient agentic image-generation workflows.

💬 Zetta ζ — A paper on a closed-loop harness for evolving runtime critics and recovery skills in physical intelligence.

📚 Read — 4 items

📚 Asana cleared 5 years of engineering work in 2 weeks with Codex — Asana reports clearing 5 years of engineering backlog in 2 weeks with Codex — a production-scale case study with concrete throughput numbers for anyone evaluating coding agent ROI.

📚 Cross-Model Memory Transfer via Target-Side Reader Adaptation — Cross-model memory transfer works by adapting a lightweight target-side reader, not the memory tables themselves — relevant for swapping models without losing agent memory.

📚 The new GitHub Copilot experience in Slack — GitHub Copilot lands in Slack with @GitHub for code review and multiplayer workflows — a new surface for teams already living in Slack.

📚 New in Air: Claude Subscriptions, Multiproject View, and Improved Markdown — JetBrains Air adds Claude subscriptions, multiproject view, and improved markdown — operator-relevant for teams managing multiple projects and model subscriptions.

Real production numbers (Asana) and memory-transfer research both point the same way: coding agents are becoming operational infrastructure, not experiments.

⚙️ Ship — 2 items

📦 Anthropic SDK 0.33.0 — Files and Skills APIs hit GA; computer use and browser use toolsets land in the TypeScript SDK — operators can now ship browser/computer tooling without custom shims.

📦 Vercel AI SDK 3.1.0 — Deepgram provider: transcription options now actually sent; `diarize` no longer defaults true — paid add-on, so check your config if you relied on speaker labels.

Two SDK releases today both remove custom shims — Anthropic's Files/Skills GA and Vercel's Deepgram transcription options now actually pass through.

📚 Read — 3 items

📚 Graph Engineering in the Era of LLM Agents: From Individual Intelligence to System Intelligence — Proposes graph engineering as the coordination layer for multi-agent LLM systems — dynamic graph structures route tasks across specialized agents and manage complex workflows.

📚 Introducing the Kotlin Benchmark for AI Coding Agents — JetBrains releases a Kotlin benchmark for AI coding agents — real-world Kotlin tasks to evaluate agent performance beyond generic SWE-bench-style suites.

📚 Foundation Agents Meet Agentic Deep Research: Evidence-Grounded Clinical Code Forecasting — Agentic deep research applied to clinical code forecasting — evidence-grounded retrieval and reasoning to predict billing codes; a pattern for any domain where agents must ground outputs in retrieved evidence.

Multi-agent coordination is moving from ad-hoc orchestration to explicit graph structures; benchmarks are expanding beyond SWE-bench to language-specific real-world tasks.