📚 Read — 5 items

📚 DistilVDR: A Compact End-to-End Visual Document Retriever via Dual-Student Distillation — 524M vision-document retriever distilled from an 8B teacher — small enough to run locally in agent pipelines that index scanned docs, no relevance labels needed.

📚 SPIEval: Evaluating Large Language Models as Mobile Assistants over Scattered Personal Information — Benchmarks mobile-assistant LLMs on scattered personal data — finds major gaps in retrieval and verification, a warning for any agent that touches user context.

📚 CLAIR-Fin: An Adversarial Multi-Agent Framework for Claim-Level Verification and Adaptive Debate in Cross-Modal Financial QA — Adversarial multi-agent debate for claim-level verification in cross-modal financial QA — a fact-checking pattern that generalizes beyond finance to any high-stakes agent output.

📚 Not All Tokens Are Equal: Inflation-Aware Routing for Agentic LLM Systems — Inflation-aware routing treats tokens by value, not volume — a cost-optimization approach for agentic systems that could cut spend without dropping quality.

📚 Agentao: A Governed Local-First Runtime for Tool-Using LLM Agents — Governed local-first runtime for tool-using LLM agents — policy controls over what tools an agent can call, aimed at self-hosted deployments that need auditability.

Today's papers converge on two operator levers: cost (token routing) and control (governed runtimes). Both are solvable with current research — the gap is adoption, not capability.

⚙️ Ship — 4 items

📦 Hermes Agent 2026.8.16 — Hermes Agent v0.20.3 rolls up ~125 PRs: MCP 2.x SDK migration with stateless protocol support, bundled Bot Mode plugin, and subprocess Python runtime isolation (PYTHONHOME/PYTHONPATH). Operators on MCP should test stateless mode; Python subprocess hardening changes runtime behavior.

📦 OpenHands 1.14.0 — OpenHands 1.14.0 adds LLM pre-flight validation to catch misconfigured profiles before saving, a Git Sync page for automations, and sets Canvas default to Kimi K3 (free). Operators configuring profiles get earlier error feedback.

📦 OpenAI SDK 3.2.0 — OpenAI Python SDK 3.2.0 adds Bedrock Runtime endpoint support and shell call streaming events. Operators bridging OpenAI SDK to Bedrock get a direct path; shell call streaming improves real-time tool output.

📦 PR #124528 Gateway profile evidence — OpenClaw ships gateway profile evidence (PR #124528) — likely new audit/telemetry for gateway profiles; check release notes for what's now logged.

Protocol migrations and pre-flight validation dominate today — test stateless MCP and profile validation before rolling out.

📚 Read — 5 items

📚 ClawGym II: Exploring Black-Box RL on Agent Harness — Black-box RL framework for optimizing agents through complex harnesses — stable, scalable optimization without white-box access, aimed at long-horizon task performance.

📚 Agentic Transaction: Towards ACID-Compliant Agent Systems — ACID-style transactions for LLM agents — semantic guarantees for reliable, isolated, durable execution, moving agent workflows toward database-grade consistency.

📚 VibeWorlding: Can Multimodal Agents Construct 3D Open Worlds End-to-End? — Agentic RL framework for end-to-end 3D world construction — agents infer intent, plan scenes, invoke tools, and reflect on feedback; benchmarks expose where multimodal agents still fail.

📚 Does a Language Server Save Tokens for Coding Agents? A Measurement Methodology and Preliminary Study — Measurement study on whether language servers reduce token consumption for coding agents — preliminary results for anyone optimizing context budgets.

📚 Inducing Reward-Free Judging Rubrics that Reduce Over-Crediting in Agent Evaluation — Reward-free judging rubrics to reduce over-crediting in agent evaluation — addresses LLM judges inflating scores, a known eval failure mode.

Agent evaluation and reliability are converging: black-box RL for harness tuning, ACID transactions for execution guarantees, and rubrics to stop over-crediting. The tooling is maturing past 'does it work' to 'can we trust it'.

💬 Community — 1 item

💬 GPT-5.6 Sol Pricing Cut by 50% — OpenRouter cuts GPT-5.6 Sol pricing 50%. 177-point HN thread of operators comparing it to Claude — the throughline: capable model, far fewer thinking tokens. Real cost signal for agent workloads.

The cost floor for agent inference keeps dropping. A 50% cut on a capable, token-efficient model means operators should re-benchmark model choices quarterly, not annually.

⚙️ Ship — 2 items

📦 Hermes Agent v2026.8.18 — Hermes Agent v2026.8.18 — stable tag rolling up ~74 PRs: desktop glass/translucency surface, tabbed SESSIONS|BOTS sidebar, Bot Mode group-chat fixes, NVIDIA SkillEvaluator Tier 1 advisory scanning on skill installs, cron media-send hardening.

📦 Qwen3.8-27B-Uncensored-GGUF trending on Hugging Face — Qwen3.8-27B-Uncensored-GGUF is trending on HF — 357k downloads, 307 likes. A 27B uncensored Qwen variant in GGUF; local agent builders are pulling it.

Hermes is consolidating a big merge window into a stable tag; operators on downstream Docker/hosted deployments should pull this before the next feature wave.

💬 Community — 4 items

💬 HarnessEval-W — A paper proposing a pipeline that uses hierarchical sub-agents to verify reasoning chains in visual world evaluations.

💬 R^3-Bench — A benchmark claiming LLMs underperform when managing a shared computation budget across multiple problems.

💬 UI-Mate — An open-weight GUI agent using environment-grounded training and in-context demonstrations to increase reliability.

💬 1667 — A terminal UI for LLM-assisted fiction writing (35 points on HN).

📚 Read — 5 items

📚 How Do Agents Fail on AutoResearch: End-to-End Diagnostic Evaluation on 100 Real-World Frontier Research Tasks — 100 real-world frontier research tasks; agents fail pervasively on metacognitive self-correction and motivation, not raw capability — the gap to close before trusting autonomous research loops.

📚 Ventor-QTest: Threat-Model-Driven Verification of Vendor-Hosted LLM APIs — Black-box probes of hosted open-weight APIs measure fidelity drift under repeated and long-sequence use — useful for anyone pinning model behavior in production.

📚 Amp adds realtime voice chat with Puck — Amp adds realtime voice chat with Puck — gpt-realtime-2.1 summarizes agent responses aloud while full text lands in thread. Hands-off supervision without staring at the terminal.

📚 Cross-Disciplinary Taxonomy and Modeling of Misunderstanding Generation, Amplification, and Detection, from Pragmatics to AI Agents — Taxonomy of how misunderstandings arise, amplify, and get detected in agent interactions — useful framing for debugging multi-turn agent failures.

📚 GitHub Copilot weekly releases — August 10 — Copilot weekly changelog covers VS Code 1.133, JetBrains, CLI — scan for agent-mode changes if you run Copilot in CI.

Research day: the failure modes are metacognitive, not capability. Operators should instrument self-correction, not just benchmark task completion.

⚙️ Ship — 3 items

📦 Codex 0.148.0 — Hooks go async and can call MCP tools; `/export` saves TUI sessions to Markdown, `/status` shows thread cost, and Amazon Bedrock Runtime joins as a built-in provider.

📦 Claude platform release notes — August 18, 2026 — Workbench is now Playground in the Claude Console — supports every Messages API parameter, shows the full SDK request and API response per run, and includes templates for code execution and web search.

📦 OpenAI SDK 3.3.0 — OpenAI Python SDK 3.3.0 adds named data-residency endpoints, letting orgs pin API traffic to specific regions for compliance.

Between Codex's async hooks and OpenAI's data-residency endpoints, operators get more control over how agent traffic runs and where it lands. Worth testing before your next pipeline refactor.

📚 Read — 4 items

📚 Agentic ESOpt: Fine-Tuning Long-Horizon LLM Agents with Minimal GPU Requirements — Agentic ESOpt fine-tunes long-horizon LLM agents with evolution strategies and trajectory-level rewards — full-parameter tuning without the GPU bill. Worth a look if you're training agents on limited hardware.

📚 From Sequence to Structure: Relational Uncertainty Propagation for LLM Agents — RUPA models agent trajectories as relational graphs to propagate uncertainty across long runs — better failure detection and confidence calibration for multi-step agents.

📚 ASI-Bench: At the Dawn of Artificial Superintelligence — ASI-Bench evaluates agents on scientific exploration and innovation tasks — a benchmark for measuring how far current agents are from autonomous research.

📚 AutoMem: A Text-Gradient Recursive Self-Improvement Framework for Automated Memory Architectures Search — AutoMem uses text-gradient recursive self-improvement to automatically search memory architectures for agents — memory design as an optimization problem rather than hand-tuning.

Today's Read lane is about making agents more reliable and cheaper to train: uncertainty propagation, automated memory search, and low-GPU fine-tuning all point the same direction.

💬 Community — 1 item

💬 Devin cuts GPT-5.6 Sol pricing 70% — Devin drops GPT-5.6 Sol pricing 70% — worth checking if your Devin workloads are on that model; HN thread is light on comments so far.

Pricing pressure on coding agents is real; when a major harness cuts a model's cost 70%, it's worth re-evaluating your own per-task economics.

⚙️ Ship — 2 items

📦 OpenAI Agents SDK 0.22.0 — OpenAI Agents SDK 0.22.0 — guardrail-rejected tool output is now redacted from replay/persisted state, and non-streaming Responses with terminal `failed`/`incomplete` raise `ModelBehaviorError`. Breaking: if you pass an explicit `openai_client` with `organization`/`project`, move those to the `AsyncOpenAI` client.

📦 Anthropic SDK 0.118.0 — Anthropic SDK 0.118.0 — adds files/memory store API surface and workspace-ID helpers; session-runner retries tool-result sends for the lease TTL, and docs now warn that blocking tool bodies stall the worker heartbeat.

Runtime-hardening day: both SDKs tighten state handling and provider contracts — worth a quick migration check if you pass explicit clients or rely on replayable traces.

💬 Community — 3 items

💬 LEGO-RL: Harness-Native Reinforcement Learning for Coding Agents — Open-source framework for harness-native RL on coding agents — in-process LLM proxying and sandbox orchestration let you policy-gradient-train the agent inside its actual harness, not a stripped-down eval loop.

💬 Agent Lightning v1.0: Towards Harnessed Agentic RL — Reproducible RL for arbitrary agent harnesses — trains the full agent stack (harness + LLM) and reports substantial coding-agent gains. Worth a look if you've been RL-tuning models in isolation.

💬 StartupBench: Benchmarking General-Purpose Agents on Market-Validated End-to-End Workflows — Benchmarks general-purpose agents on market-validated startup workflows end-to-end; even top models finish only ~30% of tasks. A reality check on how far agents are from replacing a junior operator.

The RL-for-agents field is converging on harness-native training. If you're building coding agents, the question is no longer whether to RL-tune, but which harness integration you'll adopt.

📚 Read — 4 items

📚 HarnessRisk: A Lifecycle-Oriented Benchmark for Agent Harness Safety — Benchmarks agent harness safety across six operational phases — configuration vulnerabilities and detection gaps are the weak points, not model behavior.

📚 MCP in Orbs — Amp adds remote MCP server support across orbs, the TUI, and Puck — Streamable HTTP with OAuth. Connect hosted servers from ampcode.com/settings/mcp-servers.

📚 Replit expands access to software creation with GPT-5.6 Luna — Replit adds GPT-5.6 Luna as a model option for software creation — a new capability for Replit Agent users.

📚 Cross-Model Memory Transfer via Target-Side Reader Adaptation — Cross-model memory transfer via target-side reader adaptation — lets a model inherit memory from another without retraining, relevant for agent memory portability.

Harness safety and MCP plumbing are the day's themes — operators should audit config surfaces and expect MCP support as default.

⚙️ Ship — 4 items

📦 openclaw release-publish/abd63c06c2f5-1786646045 — New release.

📦 Codex 0.147.0 — New Codex release.

📦 Gemini CLI 0.56.0 — New Gemini release.

📦 Claude Platform — Files API and Enterprise Admin API endpoints are now generally available.

💬 Community — 3 items

💬 SkillForge: Self-Distilling Agents for Project-Specific Issue Resolution — SkillForge lets agents self-distill repo-specific skills by synthesizing issues, then reusing them to resolve real ones — a step toward agents that onboard themselves to a codebase.

💬 Demystifying Agent Skills: Why They Work-Until They Don't — Skills work by anchoring execution procedurally, not by adding knowledge — and they break when the procedure doesn't match the task. Useful for deciding when to encode a skill vs. rely on the model.

💬 Harness the Memory: A Holistic Evaluation of Memory Substrates in Memory Agents — Controlled eval of memory substrates for long-horizon agents shows no single substrate wins — trade-offs depend on task regime, pushing toward adaptive substrate selection.

Today's papers converge on a theme: agents get better by learning their environment (repo skills, memory substrates), but the gains are regime-dependent — test before you standardize.

📚 Read — 5 items

📚 Towards Safer RAG: Only Agents Capable of System 2 Thinking may Access Untrusted Documents — Proposes gating untrusted document access to agents capable of System 2 reasoning, arguing fast pattern-matching agents are vulnerable to prompt injection in RAG pipelines.

📚 KernelArc: A Multi-Agent Framework for GPU Kernel Optimization — Multi-agent framework where specialized agents collaborate on GPU kernel optimization; reports speedups over single-agent baselines and existing autotuners.

📚 SkillEffect: Checked Lowering for Memory-Bounded Agent Tools — Introduces checked lowering for agent tools, enforcing memory bounds at compile time to prevent unbounded memory growth in long-running agents.

📚 GxP-Agent: Process-DAG Topology for Reliable Clinical Trial Programming with LLM Agents — Structures LLM agents with process-DAG topology for clinical trial programming, targeting reliability in regulated GxP workflows.

📚 aDSL: Agentic 3D Creation via Joint Agent-Program Design — Co-designs a DSL with a multi-agent system for 3D program synthesis; relational operators let agents compose geometry instead of emitting raw code.

Research is converging on guardrails for agent tool use: memory bounds, reasoning gates, and structured topologies are the new reliability primitives.

⚙️ Ship — 2 items

📦 Claude Files API and Admin API user management GA — Claude Files API and Enterprise Admin user-management endpoints go GA — beta headers no longer required; file expiration and pagination are now standard.

📦 Anthropic SDK 0.120.0 — Adds managed agents web search config and self-hosted sandbox memory — new knobs for agent search and sandbox persistence.

Platform maturity day: Claude's Files and Admin APIs drop beta headers while the SDK adds managed-agent search and sandbox memory. Operators can start removing beta flags.

💬 Community — 2 items

💬 Claude Code automates neuroscientific data analysis — Eli Sennesh: Claude Code automates most of his one-time job as a neuroscientific data analyst — "which is good, because I super hated that this was an entire job." 61 likes.

💬 Show HN: HarnessRouter: Unified interface for agent harnesses — Canonical API for running Codex, Claude Code, Hermes, and other managed agent harnesses as your product backend. 10 points, 14 comments.

Harness abstraction is the theme: a unified API across Codex/Claude Code/Hermes, and Claude Code already replacing a whole analyst job.

📚 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.