Morning Brief: Wednesday, July 15

Sixty-five feeds. Two weeks. 4,553 items reduced to what follows. (what we track, how we crawl, subscribe)

Wednesday morning. The Codex frame from Tuesday's Latent Space post chips in as a third-day pattern — HN files Solving 20 Erdős Problems with 20 Codex Accounts Running in Parallel, the direct follow-through on Tuesday's Codex usage up >10x to 7M users. Anthropic Claude gets its own headline: HN's I tricked Claude into leaking your deepest, darkest secrets (the memory-heist post) is Wednesday's first Claude-native red-team writeup. Tailscale ships TS-2026-009 — insecure argument handling in Tailscale SSH permitted root access. Nature Machine Intelligence's cover-package continues into day 3 — Enabling local neural operators to perform equation-free system-level analysis. Murat Demirbas ("MuratBuffalo") announces Leaving Buffalo: A Move-ing Story. House votes for permanent Daylight Saving Time. Iran mobile-network vulnerabilities story travels from Tuesday TC/Slashdot into Wednesday HN. Wednesday-morning arXiv OAI-PMH cursor has not yet refreshed — no fresh 2607.* IDs on the morning pass; expect the drop in the afternoon crawl.

Top (5-7 min)

Solving 20 Erdős Problems with 20 Codex Accounts Running in Parallel: Hacker News, 2026-07-15. The direct Wednesday follow-through on Tuesday's Latent Space Codex usage up >10x to 7M users. The Codex-parallel-agents frame is now visibly the third-day pattern; the object here is a hobby math benchmark, but the run N Codex accounts in parallel to brute-force a research task shape is the signal.
I tricked Claude into leaking your deepest, darkest secrets: Hacker News, 2026-07-15. First Claude-native red-team writeup of the week on HN — the memory heist. Pair with Tuesday's ANCHOR CLI-agent auditing paper and the Agents Don't Just Agree, They Remember persistent-sycophancy arXiv work. Memory-as-attack-surface is a coherent Tuesday→Wednesday thread.
TS-2026-009: Insecure argument handling in Tailscale SSH permitted root access: Hacker News, 2026-07-15. Tailscale SSH root-access CVE. The Wednesday morning security headline; concrete counterweight to the memory heist post as real CVE vs. red-team demo.
Enabling local neural operators to perform equation-free system-level analysis: Nature Machine Intelligence, 2026-07-15. NMI's cover-package rolls into day 3. Monday: brain-diverse computing, sustainability robotics, embodied intelligence (ergoCub). Tuesday: superposition→sparse codes, SDG realignment. Wednesday: local neural operators. A distinct three-day publication cadence.
Leaving Buffalo: A Move-ing Story: Murat Demirbas, 2026-07-15. Notable distributed-systems academic (paxos/BFT primary reader) announces career move. Pair with Hillel Wayne's Monday I'm still alive — two long-running formal-methods/distributed-systems bloggers publishing personal posts within 48h.
Iran Abused Mobile Networks' Vulnerabilities To Locate US Military In Middle East: Slashdot, 2026-07-15. The Tuesday TechCrunch report travels into Wednesday Slashdot. SS7/mobile-network-signaling attack surface is periodically rediscovered as an intelligence primitive; this week the vector is state-actor targeting of US personnel.
House Votes For Permanent Daylight Saving Time: Slashdot, 2026-07-15. Timezone-database consequences are already the developer-visible angle. Watch for tzdata proposals, calendar-app bug reports, and the recurring pilots and doctors argue against discourse.

Themes this week

Codex-as-parallel-agents (third-day pattern): HN: Solving 20 Erdős Problems with 20 Codex Accounts Running in Parallel (today), Latent Space: Codex usage up >10x to 7M users — did Codex overtake Claude Code? (Tue), TNS: OpenAI hits 8 million Codex users — what developers need to know (Tue).
Claude memory as attack surface: HN: I tricked Claude into leaking your deepest, darkest secrets (today), arXiv: ANCHOR: Automated Alignment Auditing for CLI Agents on Real-World Harm (Tue), arXiv: Agents Don't Just Agree, They Remember: Benchmarking Persistent Sycophancy in Stateful Personal Agents (Tue).
Nature Machine Intelligence cover package (day 3): NMI: Enabling local neural operators to perform equation-free system-level analysis (today), NMI: From neural superposition to sparse interpretable codes (Tue), NMI: Realigning AI technology towards the Sustainable Development Goals (Tue), NMI: The brain is a diverse place, why not computing? (Mon), NMI: Towards shared embodied intelligence in humanoid robots — ergoCub (Mon), NMI: A manifesto for Sustainability Robotics (Mon).
AI capex-cycle stress signals hardening: TC: DeepSeek reportedly in talks to raise $1.5B, then IPO (Tue), TC: Meta's Adam Mosseri — AI token budgets could soon be capped per engineer (Tue), Slashdot: IBM Stock Collapses After a Grave Warning About AI (Tue), Slashdot: 200 Economists — we must act now on AI's economic impact (Tue).
Non-transformer / non-x86 architecture signals: HN: RISC-V Is Inevitable — State of the Union Keynote Argues (today), MIT TR: PsiQuantum's photonic massive-quantum plan (Tue), HN: Is x86 ready to ACE it? (Tue), HN: Alternative(s) to run CUDA on non-Nvidia hardware — Spectral Compute (Tue).
Long-running-blogger personal posts week: Murat Demirbas: Leaving Buffalo (today), Hillel Wayne: I'm still alive (Tue).

Scan (15 min)

Wednesday signal (short)
- Solving 20 Erdős Problems with 20 Codex Accounts Running in Parallel, HN, 07-15
- I tricked Claude into leaking your deepest, darkest secrets, HN, 07-15
- TS-2026-009: Insecure argument handling in Tailscale SSH permitted root access, HN, 07-15
- Enabling local neural operators to perform equation-free system-level analysis, NMI, 07-15
- Leaving Buffalo: A Move-ing Story, Murat Demirbas, 07-15
- Iran Abused Mobile Networks' Vulnerabilities To Locate US Military In Middle East, Slashdot, 07-15
- House Votes For Permanent Daylight Saving Time, Slashdot, 07-15
- RISC-V Is Inevitable: State of the Union Keynote Argues, HN, 07-15
HN Wednesday tabs
- Jurassic Park computers in excruciating detail, HN, 07-15
- Vancouver PD website features Quick Escape button that wipes itself from history, HN, 07-15
- America pays workers just 27% of what its wealth allows — the worst in the OECD, HN, 07-15
- Who's running all those tiny RPKI servers?, HN, 07-15
TechCrunch Tuesday-into-Wednesday (VC/product beat)
- OpenAI researcher Miles Wang in talks to launch AI drug discovery startup valued at $2B, TechCrunch, 07-15
- DeepSeek reportedly in talks to raise $1.5B, then IPO, TechCrunch, 07-14
- Meta's Adam Mosseri says AI token budgets could soon be capped per engineer, TechCrunch, 07-14
- OpenAI's new flagship model deletes files on its own, people keep warning, TechCrunch, 07-14
- OpenAI's first hardware device is reportedly a screenless speaker that can move, TechCrunch, 07-14
- Anthropic's newest ad is creeping people out, TechCrunch, 07-14
Hackaday Wednesday
- Hacking Around the Financial Pain of New 3DS XL Top Screens, Hackaday, 07-15
- AIM-ing For a More Open Platform Than Discord, Hackaday, 07-15
- Benchmarking Repairability Scores with an Asus Tablet, Hackaday, 07-15
Corp-eng Wednesday
- How Bullet uses ClickHouse Cloud to give DeFi's fastest exchange real-time analytics, ClickHouse, 07-15

Tail

Faculty Job Search Series: Teaching Statements, MIT calendar, 07-15
The Gordon Institute: Virtual Information Session, Northeastern events, 07-15
Swan Boats & Picnic in the Public Garden, Northeastern events, 07-15

Feed silences (diagnostic)

arxiv-cs-ai: 0 items in the warehouse for 2026-07-15 as of the morning crawl. OAI-PMH resumption cursor has not yet advanced past Tuesday's 2607.* drop; the Wednesday fresh set will land in the afternoon crawl. Not a crawler failure — this is the OAI-PMH cadence interacting with the 10:00Z crawl time.
Anthropic first-party, OpenAI, Apple ML Research, Google Research, EFF Deeplinks, Schneier, GitHub Blog, Interconnects, Kenneth Payne, Martin Fowler, Microsoft Research, Cloudflare, Charity Majors, Databricks, Hugging Face Blog, Alignment Forum, Quanta Magazine, Grafana Labs, 404 Media, Pluralistic, Pinboard, Andrej Bauer, LWN, WBUR, Simon Willison, Terence Tao, Latent Space, MIT Technology Review, TheNewStack, Vercel, Planet Clojure, claude-code-releases: no primary posts yet on Wednesday morning (silence is expected — most publish Tuesday afternoon US/EU or Wednesday midday).
Netflix Tech Blog, James Bornholt: persistent DNS/TLS errors continue (unchanged from Sun/Mon/Tue).
bitsavers (6 feeds): connected, 0 items (sparse archive, expected).

Build provenance

build: 2026-07-15 | crawler-sha: 5fe7ab8 (Walsh-Research/1.2, compliance v1.3) | feeds: 65 active (78 configured, incl. 12 corp-eng, 6 bitsavers, 5 generated) | items-considered: 4553 (14d, incl. 2793 arXiv) | warehouse: 25407 items | published: 7 | note: Wednesday morning. Codex-as-parallel-agents chips in as a third-day pattern — HN files "Solving 20 Erdős Problems with 20 Codex Accounts Running in Parallel," direct follow-through on Tuesday's Latent Space Codex-vs-Claude-Code post. Anthropic Claude memory-heist writeup lands ("I tricked Claude into leaking your deepest, darkest secrets"). Tailscale ships TS-2026-009 (SSH root-access CVE). NMI's cover-package continues into day 3 (local neural operators for equation-free system-level analysis). Murat Demirbas announces leaving Buffalo. House votes for permanent Daylight Saving Time. Iran mobile-network vulnerabilities story travels from Tuesday TC/Slashdot into Wednesday HN. arXiv cs.AI cursor has not yet advanced on the morning pass — fresh 2607.* drop expected in the afternoon crawl.