Morning Brief: Friday, October 23
Seventy-four feeds reporting. Two weeks. 6,104 items reduced to what follows. (what we track, how we crawl, subscribe)
Unvetted agents are already reaching into US bank accounts. The same morning's research argues the audits meant to catch them are formally bounded.
Three papers landed in today's cs.AI batch making versions of the same negative claim, and they did not coordinate. Post-Audit Mirage shows that offline agent verification hits an identifiability wall: past a point, the trace you kept cannot distinguish the policy you have from one you would reject. What Lies Beyond Perception gives the partition argument for why evaluation and oversight have a ceiling at all. What the Trace Shows and What the Label Says turns the instrument on itself and audits the behavioral detectors. Read them against the day's other half — a Three-Pillar Governance Framework, a Petri-net Certified Execution Harness, the VERA and TESSA EU AI Act dashboards, AIRO risk ontologies — all of which assume the audit works. METR's Thursday red-team of OpenAI's internal agent monitoring is the empirical version of the same doubt.
php2shell entered week two as operations rather than news. InfoWorld's
post-mortem argues the two-month embargo is the structural defect, not
the bug; Schneier and Zoe Kleinman supply the "unexploded bomb"
framing; and someone posted a genuinely sad Tell HN apologizing
for unknowingly breaking the embargo. The detail that belongs in a
crawler brief: The Crawler Zoo catalogued a self-declared arrival
today calling itself wp2shell-rce. Different target — WordPress, not
core PHP — but the same naming convention, and a user agent that
announces its own exploit intent in the string.
Top (5-7 min)
- Unvetted AI agents are accessing US bank accounts
- Slashdot, 2026-10-23. The deployment fact the rest of today's research is reacting to. Agents reached regulated systems before the assurance machinery did.
- Post-Audit Mirage: Identifiability Limits of Offline Agent Verification
- arXiv cs.AI, 2026-10-23. An identifiability result, not a benchmark. If it holds, some agent verification questions are not answerable from logs at any budget.
- What Lies Beyond Perception: A Partition Theory of the Limits of AI Evaluation and Oversight
- arXiv cs.AI, 2026-10-23. The general form of the same ceiling. Theory-first and worth the effort before you buy an oversight product.
- A Three-Pillar Governance Framework for Safe and Trustworthy AI Agents
- arXiv cs.AI, 2026-10-23. The constructive half of the day. Read it knowing two of the morning's other papers bound what its audit pillar can establish.
- Red-teaming OpenAI's internal agent monitoring systems
- METR, 2026-10-22. Someone actually attacked a production monitor. The empirical counterpart to the theory above.
- Crossing the threshold: how OpenAI and Anthropic approach autonomy and safety in financial agents
- Latent Space, 2026-10-14. The labs' own framing of exactly the deployment Slashdot is reporting. Arrived late in today's backfill; read it against the bank-account item above.
- PHP2Shell exposes the weakness in coordinated vulnerability disclosure
- InfoWorld, 2026-10-23. The argument that the disclosure process, not the vulnerability, is what needs redesigning now that discovery is cheap.
- New arrival: wp2shell-rce
- The Crawler Zoo, 2026-10-23. A self-declared bot whose user agent names the exploit it is hunting. One of six new arrivals today.
Themes
The oversight you can buy is bounded
- What the Trace Shows and What the Label Says: Auditing Behavioral Detectors in Agent Evaluation
- arXiv cs.AI, 2026-10-23. Audits the auditors. The gap between what a detector fires on and what its label claims is the finding.
- Faithful latent harm monitors work no better than the model's own behaviour at a matched false alarm rate
- arXiv cs.AI, 2026-10-23. A negative result stated at matched FAR, which is the only way this comparison means anything.
- Scaling Laws for Safety Probes
- arXiv cs.AI, 2026-10-23. Asks whether probe quality scales the way capability does. The answer determines whether monitoring keeps pace.
- Simulated Control Evals
- arXiv cs.AI, 2026-10-23.
- Monitorability evaluations and policy proposals
- Alignment Forum, 2026-10-22. The policy side of monitorability, written before this week's limit results. Worth revisiting against them.
- Who Is Sycophantic to Whom? Confident LLMs, Human Deference, and the Collapse of Trust Calibration
- arXiv cs.AI, 2026-10-23. Oversight failure located in the human half of the loop.
- Probing and Steering for Safety-Relevant Behaviours in Large Language Models
- arXiv cs.AI, 2026-10-23.
- Questioning Instrumental Convergence in RLVR: Checkpoint Evidence from OLMo-3
- arXiv cs.AI, 2026-10-23. Checkpoint-level evidence against a widely assumed dynamic. Open weights make the claim checkable.
Governance, shipped as software
- Superintelligent Software Engineering: Governance That Scales with Capability
- arXiv cs.AI, 2026-10-23. Treats governance as something that has to scale, not a fixed checklist. The framing is the contribution.
- Certified Execution Harness for Agentic AI: Safe and Least-Intrusive Petri-Net Supervisory Control Between Reasoning and Action
- arXiv cs.AI, 2026-10-23. Supervisory control theory applied at the reason/act boundary. The most concretely implementable thing in this section.
- VERA: A Responsible AI Dashboard for Assessing LLM Deployments in Light of the EU AI Act
- arXiv cs.AI, 2026-10-23.
- TESSA: A Governance-Oriented Custom GPT for SME Guidance under the EU AI Act
- arXiv cs.AI, 2026-10-23. Compliance guidance delivered as a chatbot. Note the recursion before adopting it.
- AI Risk Ontologies for the EU AI Act Compliance: AIRO, VAIR, and their Integration with DPV
- arXiv cs.AI, 2026-10-23.
- Contestable Agents for Data Governance: Recording Human Stewardship of AI-Generated Metadata
- arXiv cs.AI, 2026-10-23.
- RIDAR: Safeguarding Research Integrity in Autonomous Research Agents via Automated Detection and Runtime Intervention
- arXiv cs.AI, 2026-10-23.
- The Delegation Divide: An Open Index of AI-Agent Delegability across Occupations and Countries
- arXiv cs.AI, 2026-10-23. An open index rather than an op-ed about job loss. Useful as data even if you dispute the scoring.
- Governed Self-Evolution for Autonomous Investing: A Multi-Agent LLM System with an Invariant Core
- arXiv cs.AI, 2026-10-23.
- Automated screening of privacy risks in medical AI
- arXiv cs.AI, 2026-10-23.
- Localizing compliance at scale: achieving South Korea's ISMS certification
- Cloudflare, 2026-10-22. What jurisdiction-by-jurisdiction compliance actually costs an infrastructure operator. The unglamorous version of the above.
Agent security
- AgentPhishing: A Benchmark of LLM Email Agents Under Adaptive Attack
- arXiv cs.AI, 2026-10-23. Adaptive, not static, attacks against email agents. The adaptivity is what makes the numbers worth anything.
- Agent Mosaic: Recognizing Confidentiality Leaks in Agentic LLM Systems
- arXiv cs.AI, 2026-10-23.
- AgentTriage: Replayable Security Diagnosis Across Heterogeneous Agent Runtimes
- arXiv cs.AI, 2026-10-23. Replayability as a security primitive. Compare against the offline verification limits in the Top section.
- Governance-State Corruption in Long-Term Memory for LLM Agents: A Cross-Model Stress Test
- arXiv cs.AI, 2026-10-23. Where the memory thread and the governance thread meet: the policy state itself is writable.
- The Price of Safety: Benign-Case Utility and Token Overhead of Memory-Poisoning Defenses in LLM Agents
- arXiv cs.AI, 2026-10-23.
- MPIDetector: Detecting Multi-modal Prompt Injection via LLM-based Reasoning
- arXiv cs.AI, 2026-10-23.
- LLM Multi-Agent Systems Are Vulnerable to Herding and Runaway Cascades
- arXiv cs.AI, 2026-10-23. Failure mode specific to multi-agent setups, with no single compromised component to find afterwards.
- Measuring the Permission Gate: A Stress-Test Evaluation of Claude Code's Auto Mode
- arXiv cs.AI, 2026-10-23. Someone stress-tested a shipped permission gate and published the numbers. Directly relevant if you run agents with tool access.
- SecLog-Bench: A Scalable Benchmark for Cybersecurity Log Reasoning with Scaffolded Evidence
- arXiv cs.AI, 2026-10-23.
- Secure Agentic AI and Multimodal Intelligence in 6G Communication Systems
- arXiv cs.AI, 2026-10-23.
php2shell, week two
- New serious PHP vulnerabilities released and patched
- Schneier on Security, 2026-10-22.
- Zoe Kleinman: PHP is the cyber equivalent of an unexploded bomb
- Slashdot, 2026-10-22.
- Tell HN: I unknowingly broke the PHP RCE embargo. I am sorry.
- Hacker News, 2026-10-22. The human cost of a two-month embargo, told by someone who tripped over it. Read it alongside InfoWorld's structural argument.
- AI is changing everything about hacking. No one knows what happens next.
- 404 Media, 2026-10-22.
- GitLab critical patches include one for a bug that allows code execution by unauthenticated users
- InfoWorld, 2026-10-22.
- CVE-2026-111943: compromised intermediate certificate authority
- Hacker News, 2026-10-22.
- eMudhra to exit global SSL/TLS certificate business
- Hacker News, 2026-10-22. A CA leaving the business the same week another intermediate turns up compromised. The timing is the story.
- Several Linux kernel vulnerabilities accidentally posted to public kernel mailing list
- Slashdot, 2026-10-22.
- Header Accordions: QPACK amplification attacks against HTTP/3 servers
- Lobsters, 2026-10-22.
- How bug bounty sharpens one researcher's security skills
- GitHub Blog, 2026-10-22.
- LLMs are making formal software verification feasible at a much larger scale
- Hacker News, 2026-10-22. The optimistic reading of the same capability. Verification, not discovery, is the claim being made.
Agent memory
- The Value of Memory in Language Agents
- arXiv cs.AI, 2026-10-23. Tries to price memory rather than assume it. The right question to ask before adding a store.
- Is complex agentic memory always needed? Naive RAG is stronger than we thought
- arXiv cs.AI, 2026-10-23. The deflationary result. Nine memory papers in one batch and this is the one arguing the baseline already works.
- ColaBench: Evaluating Memory-Guided Knowledge Acquisition in Personal Agents
- arXiv cs.AI, 2026-10-23.
- Why Bother Me Again? Benchmarking Personal Agent Memory for Interaction Preferences
- arXiv cs.AI, 2026-10-23.
- Memory-Aware Reward Modeling: Learning Human Preferences for LLM Memory Use
- arXiv cs.AI, 2026-10-23.
- From Faulty Memories to Corrected Actions: Dependency-Guided Rollback Repair for Memory-Augmented Agents
- arXiv cs.AI, 2026-10-23.
- When LLMs Forget What They Believe: Characterizing State Integrity Collapse in Long-Horizon Dialogues
- arXiv cs.AI, 2026-10-23. State integrity as a measurable property of long sessions. Useful framing for anyone running multi-hour agent loops.
- Externalise the Memory: A Pre-Registered Study of Where an Embodied Agent's Experience Should Live
- arXiv cs.AI, 2026-10-23. Pre-registered, which is rare enough here to note on its own.
- SpatialMem: Metric-Aligned Long-Horizon Video Memory for Language Grounding and QA
- arXiv cs.AI, 2026-10-23.
Coding agents and harnesses
- Can LLMs Improve Their Own Harness? A Controlled Study on ARC-AGI-3
- arXiv cs.AI, 2026-10-23. A controlled study of the self-improvement claim, scoped to the harness rather than the weights. The scoping is what makes it answerable.
- Agentic Verified Optimization of Real-World Software
- arXiv cs.AI, 2026-10-23.
- Automating Feature Flag Cleanup at Scale with Multi-Agent LLM Systems: An Industrial Experience at DoorDash
- arXiv cs.AI, 2026-10-23. An industrial report on a boring, well-scoped task. More informative than most capability papers.
- SWE-VERIFIER: Selective Verification for Cost-Efficient Coding Agents
- arXiv cs.AI, 2026-10-23.
- RIPPLE: Recovery Internalization through Process-Level Policy Learning and Effect-Aware Credit for Long-Horizon Agentic Coding
- arXiv cs.AI, 2026-10-23.
- Adaptive Skill Granularity for Coding Agents: Runtime Routing and Macro-to-Micro Recovery
- arXiv cs.AI, 2026-10-23.
- MemMux: Runtime Verification and Honest Resource Attribution for Fleets of Parallel Coding Agents
- arXiv cs.AI, 2026-10-23. Resource attribution across parallel agents, which is the billing problem nobody wants to own yet.
- Self-Evolving Static Analysis via LLM Pseudo-Labeling
- arXiv cs.AI, 2026-10-23.
- LLM2Cert: LLM Synthesis of Certificates with SMT/Lean Verification
- arXiv cs.AI, 2026-10-23. Machine-checked output rather than a judge's opinion. The pattern worth copying from this batch.
- Beyond LLM Judge: Mechanical Verification for Agentic Autoformalization
- arXiv cs.AI, 2026-10-23.
- MCP vs CLI for AI Agents: A Comparative Analysis of Tool Integration Strategies for Autonomous Systems
- arXiv cs.AI, 2026-10-23. Somebody finally measured the argument instead of having it on social media.
- The Cost of Saying Hello: An Empirical Study of Handshake Overhead in AI Agent Communication Protocols
- arXiv cs.AI, 2026-10-23.
- Run Pydantic AI coding agents in E2B sandboxes
- Pydantic, 2026-10-22.
- Give your Supabase app its own MCP server
- Supabase, 2026-10-22.
- Trace and evaluate Cursor agents with Langfuse, powered by ClickHouse
- ClickHouse, 2026-10-22.
- Donating Agent Skills to the Agentic AI Foundation
- Simon Willison, 2026-10-22. A format moving to a foundation. Relevant to anyone who has written skills against it.
- Claude Mods: flipping the harness inside out
- Latent Space, 2026-10-21. Arrived in today's Latent Space backfill, on-theme with the harness papers above.
- Left alone with nothing to do, AI agents started playing
- TechCrunch, 2026-10-22.
Evaluation and judges
- Behind Aggregate Scores: What Systematically Varying Arena-Hard's LLM Judge Reveals
- arXiv cs.AI, 2026-10-23. Varies the judge on a leaderboard everyone cites. The sensitivity is the result.
- Prompting Strategy Effects Depend on the Judge: A Multilingual Study of LLM-as-a-Judge Robustness
- arXiv cs.AI, 2026-10-23.
- What Pairwise Evaluation Misses: Reward Scores across Prompts
- arXiv cs.AI, 2026-10-23.
- Scaling Laws for Generative Reward Models
- arXiv cs.AI, 2026-10-23.
- Major Revision, Minor Accountability: LLMs as Reviewers of Published Articles in Top-Tier Journals
- arXiv cs.AI, 2026-10-23.
- FLAWS: A Benchmark for Error Identification and Localization in Scientific Papers
- arXiv cs.AI, 2026-10-23.
- Stop Drawing Scientific Claims from LLM Social Simulations Without Robustness Audits
- arXiv cs.AI, 2026-10-23. A title that is also the abstract. The robustness audit is the ask.
- Prompt Sensitivity in Tool-Augmented Model Comparisons
- arXiv cs.AI, 2026-10-23.
- Predictable performance is suboptimal, and what it means for model routing
- arXiv cs.AI, 2026-10-23. Argues variance is worth paying for in a router. Counterintuitive and cleanly stated.
- Debunking "faster than Rust": malicious benchmarks in Tov
- Lobsters, 2026-10-23. The non-AI version of the same problem. Benchmarks are an adversarial medium wherever the claim is competitive.
Infrastructure and economics
- AI's hidden storage bill: what inference leaves behind
- MIT Technology Review, 2026-10-22. The cost line nobody forecasts: what inference writes down and keeps.
- Designing an AI serving harness built to last
- MIT Technology Review, 2026-10-22.
- The new economics of AI cost management
- Databricks, 2026-10-22.
- The AI bill is coming
- Databricks, 2026-10-22.
- When resilience is out of your control
- InfoWorld, 2026-10-23.
- To make way for data centers, Google will pay others to use less power
- Slashdot, 2026-10-22. Demand response repurposed as datacenter siting policy. Watch who gets paid and who gets curtailed.
- Low-Latency MoE Inference in Space Data Centers: A Wireless-Power-Transfer Perspective
- arXiv cs.AI, 2026-10-23.
- Detecting AI Training in Datacentre Power
- arXiv cs.AI, 2026-10-23. Training runs are visible from the grid side. A measurement channel with obvious policy uses.
- Why AI compute comes in rectangles — Erik Bernhardsson, Modal CEO
- Latent Space, 2026-10-22.
- Build more responsive apps with the Decisions API
- OpenAI, 2026-10-22. Same 403.
- Single-tenant Titus: VM-level isolation on a multi-tenant container platform
- Netflix Tech Blog, 2026-10-22. Feed-provenanced; Medium returns 403 to our verifier. Netflix breaking a six-day silence with real systems writing. Relevant if you are isolating untrusted agent workloads.
Policy, labor, courts
- Google staff resign in backlash over its AI contracts with military
- Slashdot, 2026-10-22.
- AI is already 'amplifying the destruction' of war
- 404 Media, 2026-10-22.
- How generative AI is scaling transnational repression
- Citizen Lab, 2026-10-22. Citizen Lab documenting the deployment, not speculating about it.
- Canadians are running out of time to oppose the dangerous C-22 surveillance bill
- EFF Deeplinks, 2026-10-22.
- Portal Texas used to 'rubber stamp' data centers down after 404 Media reporting
- 404 Media, 2026-10-22.
- OpenAI and Poland's NASK partner to advance AI safety and security
- OpenAI, 2026-10-23.
openai.comreturns 403 to our verifier. - UN says breaching 1.5C warming limit is 'imminent and unavoidable'
- Slashdot, 2026-10-23.
- Gender, Sexual Orientation, and the One-Year Refugee Bar: A Computational Study of Bill C-12 and Canadian Refugee Adjudication
- arXiv cs.AI, 2026-10-23.
- Harm Is Behavioural, Moderation Is Lexical: How Language and Culture Decide Whom Automated Safety Systems Protect and Whom They Silence
- arXiv cs.AI, 2026-10-23. A clean statement of the mismatch at the centre of most moderation failures.
- In the biotech world, you can't see everything
- MIT Technology Review, 2026-10-23.
- Moderna shares plunge after leak of cancer vaccine data
- Slashdot, 2026-10-22.
Systems
- The end of __get_user() and __put_user()
- LWN, 2026-10-22. Subscriber link;
lwn.netblocks our verifier (403). - Rust features in the future
- LWN, 2026-10-22. Subscriber link; same 403.
- KeePassXC 2.8.0 released
- LWN, 2026-10-22.
- UUIDv7 doesn't have the properties you think it does
- Lobsters, 2026-10-22. Read it before you rely on v7 monotonicity across processes.
- Objects, attributes, maps, oh my
- Lobsters, 2026-10-22.
- I built a garbage collector for Rust (2023)
- Lobsters, 2026-10-22.
- Script dependencies with Nix
- Lobsters, 2026-10-22.
- A hacky way to portably access errno from outside of C
- Lobsters, 2026-10-22.
- Expose only share links of a private service with Caddy
- Lobsters, 2026-10-22.
- Postgres 19 is late, and it's AI's fault
- Neon, 2026-10-21. A Postgres vendor putting a release slip down to AI. Broke a six-day Neon silence in today's crawl.
- Docker is weird
- Hacker News, 2026-10-22.
- Jank: the native Clojure dialect with seamless C++ interop
- Hacker News, 2026-10-22.
- Clojure 1.13.0-alpha10
- Planet Clojure, 2026-10-22. The releases feed broke an eleven-day silence to say so.
- slk 0.30.0
- Planet Clojure, 2026-10-22.
- editmd 0.1.0
- Planet Clojure, 2026-10-23.
- Flint-interactive: a language for creating interactive data visualizations
- Microsoft Research, 2026-10-22.
- We have traits at home
- Lobsters, 2026-10-22.
- The Hibernate origin story (outsider edition)
- Lobsters, 2026-10-23.
- 686: Yell in your Shell
- BSD Now, 2026-10-22.
- NYC, new webinar, differential testing
- Hillel Wayne, 2026-10-22. First item in weeks. Differential testing is the technique most adjacent to this week's verification papers.
Crawlers and the open web
- New arrival: ev-crawler
- The Crawler Zoo, 2026-10-23. Six new crawlers catalogued today, 120 across the fourteen-day window.
- New arrival: TriumviryResearchBot
- The Crawler Zoo, 2026-10-23.
- New arrival: IndexaroLeadBot
- The Crawler Zoo, 2026-10-23.
- New arrival: AwarioBot
- The Crawler Zoo, 2026-10-23.
- New arrival: AdsBot-Google
- The Crawler Zoo, 2026-10-23.
- Please don't mess with links
- Lobsters, 2026-10-22.
- Posts that Hacker News liked
- Planet Clojure, 2026-10-23.
- Local Wikipedia: Wikipedia, but only for things a half hour walk from you
- Hacker News, 2026-10-22.
- Quad9 to exit France
- Hacker News, 2026-10-22. A public resolver withdrawing from a jurisdiction. The compliance cost of running infrastructure, paid in coverage.
- What is your opinion on submission quality since the vibecoding tag was removed?
- Lobsters, 2026-10-22. Feed-provenanced; lobste.rs robots.txt disallows our verifier, so the URL is unfetched by design.
Research beyond agents
- Breaking the high-degree barrier in Lehmer's problem with a human-AI discovery loop
- arXiv cs.AI, 2026-10-23. A named barrier moved, with the human-AI loop described rather than claimed. The strongest capability result in today's batch.
- An Introduction to World Models for Generalist Agents
- arXiv cs.AI, 2026-10-23.
- Dual Exam: Controlled Evaluation of Multimodal Agents and World Models
- arXiv cs.AI, 2026-10-23.
- Anticipating Social Consequences with Distributional World Models
- arXiv cs.AI, 2026-10-23.
- Trion-1: Unifying Language, Protein Sequence, and Structure with Diffusion Language Models
- arXiv cs.AI, 2026-10-23.
- Efficient equivariant framework for universal charge density prediction
- Nature Machine Intelligence, 2026-10-23.
- Guidelines for evaluating structure-based methods for protein-ligand binding prediction
- Nature Machine Intelligence, 2026-10-23. Evaluation guidelines from the field most burned by optimistic docking numbers.
- Large language models show systematic failures in incorporating evolving medical knowledge
- arXiv cs.AI, 2026-10-23. Systematic, not anecdotal. The failure is in updating, which is exactly what clinical use requires.
- The Geometry of Meaning in Neural Manifolds
- arXiv cs.AI, 2026-10-23.
- Rethinking SAE Evaluation: An Atomic Interpretable Unit Framework Reveals Hidden Polysemanticity and Redundancy
- arXiv cs.AI, 2026-10-23.
- PLE: Physical-Engineering's Last Exam
- arXiv cs.AI, 2026-10-23.
- A lost Schrödinger manuscript reveals a twist in quantum physics
- Hacker News, 2026-10-22.
- OlmoEarth v1.3: state-of-the-art embeddings for the whole planet
- Hugging Face Blog, 2026-10-22.
- Turning database grading into RL rewards: post-training agents with ThinkingBox
- Hugging Face Blog, 2026-10-22.
- Accelerating Wildfire Hazard Mapping Across Canada with an AI Surrogate Modelling Framework
- arXiv cs.AI, 2026-10-23.
- A Topology-Guided GeoAI Agentic Framework for Urban Flood Screening
- arXiv cs.AI, 2026-10-23.
Tail
- On the foolishness of "natural language programming" (1978)
- Hacker News, 2026-10-23. Dijkstra, forty-eight years ago, on why precision is the point. The best thing on the list today and the cheapest to read.
- DOOM ported to a $30 clamp meter
- Hackaday, 2026-10-23.
- Restoring a Soviet ES-4001 mainframe control panel
- Hackaday, 2026-10-23.
- I did this without your knowledge
- Hacker News, 2026-10-23.
- How does Fable 5.1 hide messages in images?
- Hacker News, 2026-10-22.
- Every AI model has the same "controversial" take on humans
- Hacker News, 2026-10-22.
- Mobile-Whip: a compact battery-powered mini-whip active HF antenna
- RTL-SDR, 2026-10-23.
- Testing Hubble Network's Bluetooth-to-satellite link from Budapest
- RTL-SDR, 2026-10-23. Bluetooth to orbit, measured by an amateur in Budapest. The implausible part is that it works.
- Jailbreaking the Switch 2
- Lobsters, 2026-10-22.
- Show HN: making a smart TV dumb and then smart
- Hacker News, 2026-10-23.
- KiCad… in the browser?
- Hackaday, 2026-10-22.
- Rule 30: one black cell
- Hacker News, 2026-10-22.
- Would an A330-900F make sense? Part 2: the cost advantage
- Leeham News, 2026-10-22.
- Oura sues rival Ultrahuman again, threatening its US smart ring comeback
- TechCrunch, 2026-10-23.
- Uber rival inDrive eyes 2027 robotaxi tests in Central Asia with Pony.ai and Nvidia
- TechCrunch, 2026-10-23.
- The Browser Company retires its Arc browser, pushes users to Dia
- Slashdot, 2026-10-22.
- Anthropic takes on Muse with Claude Duet
- Slashdot, 2026-10-22.
- Railpass
- Hacker News, 2026-10-22.
- Getting ready to operate GNU Taler in Germany
- Hacker News, 2026-10-22.
Feed silences (>72h since last item)
- Interconnects (3 days) — last item 2026-10-20.
- AI Snake Oil, Tailscale, and next.jdbc releases (4 days) — last item 2026-10-19.
- anthropic-generated (5 days) — last item 2026-10-18.
- Vicki Boykis (6 days) — last item 2026-10-17.
- cursor-blog (7 days) — last item 2026-10-16.
- Antithesis, Apple ML Research, and The Markup (8 days) — last item 2026-10-15.
- FreeBSD Foundation and Jane Street (10 days) — last item 2026-10-13.
- Babashka releases (12 days) — last item 2026-10-11.
- Boston Rust Meetup (13 days) — last item 2026-10-10.
- Charity Majors (15 days) — last item 2026-10-08.
- Marc Brooker (19 days) — last item 2026-10-04.
- nREPL releases (21 days) — last item 2026-10-02.
- deepmind-blog (22 days) — last item dated 2026-10-01, still backdated rather than absent.
Broke silence in today's run: Netflix Tech Blog (six days), Neon
(five), BSD Now (eight), Clojure releases (twelve), and
Hillel Wayne, which had been past the horizon entirely. Latent
Space also backfilled twenty-one archive items in a single poll,
dated 2026-10-10 through 2026-10-21 and all first seen today — a
re-poll artifact, not twenty-one new posts, and a reminder that
first-seen and published dates are different columns for a reason.
Fifty-four of the 118 feeds polled today returned 304 Not Modified —
conditional requests working as intended, not silence. One hard error
in today's run: James Bornholt (www.bornholt.dev does not resolve)
for the eighth consecutive day; the registry correction filed last week
is still unapplied, and this is now the longest-running known defect in
the crawl. Still beyond the
practical horizon and excluded from the table above: East Boston
Times, logback releases, Bunnie Studios, Stephen Wolfram, Ink
& Switch, TigerBeetle, Alex Ellis, Steve Yegge, Kenneth
Payne, All Things Distributed, Lambda the Ultimate, and the
dormant Clojure library release feeds.
Build provenance
build: 2026-10-23 | crawler-sha: b277e41 (Walsh-Research/1.2, compliance v1.4) | feeds: 126 registered, 118 polled today (63 ok, 54 not-modified, 1 error), 74 publishing in window | items-considered: 6104 (14d, incl. 3578 arXiv cs.AI) | warehouse: 62273 items | published: 172 (166 verified 2xx, 6 known bot-block/robots false positives)