--- title: "Daily Brief" canonical: "https://current.tinyfish.ai/issue/46/daily-brief" publication: "TinyFish" publisher: "TinyFish" language: "en-US" guide: "https://current.tinyfish.ai/llms.txt" permalink: "https://current.tinyfish.ai/issue/46/daily-brief" issue: 46 section: "daily-brief" description: "Your daily digest of web developments through the lens of enterprise agent deployment" authors: ["Sable Whitford"] published: 2026-09-14 updated: 2026-11-04 articles: 13 --- # TinyFish > CURRENT is a journal about AI agents in production, for the engineering and product leaders who have to make them work. Drawing on experience building enterprise web agents, it explains why the demo succeeded and the deployment didn't, and what that means for choosing vendors, briefing the board, and avoiding mistakes others have already made. [Read on the website]() ## Issue 46 — Daily Brief Your daily digest of web developments through the lens of enterprise agent deployment ### 30-Plus Towns Voted Down Data Centers, and the AI Buildout Has a New Bottleneck - [30-Plus Towns Voted Down Data Centers, and the AI Buildout Has a New Bottleneck]() — By [Sable Whitford]() · Published 2026-11-04 **As Wednesday's counts firm up, the AI industry is facing a bottleneck it can't fix by ordering more hardware.** In Tuesday's elections, 30-plus local measures restricting data centers passed or are leading. Ten states put 41 measures in front of voters, and exactly one, in North Dakota, went the builders' way. Granville, Ohio, banned data centers with 82.5% of the vote. Sunbury's ban lands on a planned Amazon site. Piqua threw out its mayor and two commissioners who backed a big project. In an NBC exit poll, 56% of voters called AI and data centers a major factor in their vote for Congress. Local anger that had been simmering online and at council meetings showed up at the polls. ### Agents everywhere, and the eye-watering numbers behind them While half the internet refreshes vote counts, the other half has been shipping things, filing things and getting robbed. A few of today's terms are odd even by our standards, so here's a quick decoder: - **S-1:** the filing a company makes with the SEC before it goes public. The risk-factors section legally requires it to confess everything that could go wrong, which is why some of the strangest sentences in corporate America end up there. - **Recursive self-improvement:** an AI that can build a meaningfully better version of itself. That version builds a better one, and so on. - **App Intents:** Apple's framework, around since iOS 16 in 2022, that lets apps expose actions to Siri and Shortcuts. - **Kill switch:** in ops, anything that stops a process fast *and keeps it stopped*. People reliably forget the second half. Grab a coffee first. Several of these numbers have a lot of zeros. - **Anthropic's S-1 comes with an $895B compute tab** — [Source]() — Published 2026-11-04 Tag: IPO Paperwork Nobody can stop talking about it. Q3 revenue hit nearly $17B, a 47% jump. It also posted a $567M operating loss, mostly from $5B in short-term compute. Commitments total about $895B, including up to $120.5B to SpaceX through 2029. The line everyone keeps quoting says the company is "very close" to recursive self-improvement. No model has crossed that line, and humans stay in the R&D loop. Claude wrote over 90% of September's code contributions. Skeptics call an S-1 part disclosure, part pitch deck. - **OpenAI claims ChatGPT is a fifth of Google** — [Source]() — Published 2026-11-04 Tag: Search Wars OpenAI claims ChatGPT now handles search-style queries equal to about 20% of Google's volume. It hasn't published the data behind that. When Ahrefs ran its own count in February, it came out near 12%. The company that benefits from the metric is also the one reporting it, and nobody else can check. File it under "pending." - **Siri opens up, and skills are SKILL.md** — [Source]() — Published 2026-11-04 Tag: Apple Dev The new 27.3 betas open Siri AI to every developer through App Intents, on iOS, iPadOS, macOS and visionOS. A new skills system chains several app actions into a single request. Developers are grinning about one detail: app skills are SKILL.md files, a subset of the Agent Skills format that coding agents already read. For now it only works in dev-signed builds and internal TestFlight. Customers get it in a few months, and App Store submissions open in spring. - **Amazon rewrites the software engineer job description** — [Source]() — Published 2026-11-04 Tag: Eng Culture A senior Amazon engineer says agents now write almost all the code developers used to type by hand. The memo calls counting tokens and code changes "deeply flawed" ways to measure AI use (correct, somebody frame it). Amazon also disputes reports that its Kiro agent caused an AWS outage and blames a misconfigured role instead. Every ops person alive has typed "misconfigured role" into a postmortem at least once. ### Security bills come due (and other bad mornings) - **AI is calling in 40 years of debt** — [Source]() — Published 2026-11-04 Tag: Tech Debt The day's loudest fight is over whether "security through neglect" is finished. The old bet was that ancient code stayed safe because nobody had bothered to read it. Agents will happily spend hours on code no human has opened since dial-up. The PHP2Shell patches last month fixed two critical PHP flaws. Wiz says one researcher with frontier AI did in a week what a world-class team would have needed months for. - **Binance's $6.3B heist started with a job offer** — [Source]() — Published 2026-11-04 Tag: Supply Chain Binance's incident report names two ways in. In September, a fake "project director" approached an engineer through a job-hunting platform. In June, another engineer installed a popular AI coding tool through a tampered Homebrew install. The FBI blames North Korea, and Kim Jong-un's sister calls that a "grave political provocation." Every dev in the world is now squinting at their install scripts, and at their recruiter DMs. - **CrowdStrike now sells a kill switch for agents** — [Source]() — Published 2026-11-04 Tag: Agent Ops CrowdStrike's new AI kill switch stops a rogue internal agent, blocks it from relaunching and cuts its network access. If you've ever watched systemd cheerfully restart the thing you just killed, you'll see why the relaunch block matters. CrowdStrike also announced a Claude Cowork integration and two AI acquisitions. "Agent off-switch" is officially a product category, which says plenty about how the pilots are going. - **GitHub had a morning** — [Source]() — Published 2026-11-04 Tag: Outage Watch Around 10:00 UTC today, GitHub's status page flagged an incident affecting pull requests, git operations and issues. There's no root cause yet, so put down the armchair postmortem. If your push hung or your PR wouldn't load, it wasn't you. Refill the coffee and call it a code review break. ### Favorite Featured Stories - [Who Pays While the Agent Waits]() — Published 2026-10-30 Google's new service for hosting AI agents starts from a fact anyone who has watched one run already knew: most of the time, the agent is waiting, for the model to answer or for a person to reply. So Google suspends idle agents and promises you won't pay for idle machines. Fair enough, but the wait still gets paid for. It moves into approval queues nobody times, and then onto the plan, which comes back exactly as it left, even if the stock it counted on has since sold. - [The Runnable Paper Needs an Identity Report]() — Published 2026-10-30 Paper2Agent, published in Nature in September, turns research papers into AI agents you can run. Hand over your data and the method executes, with no rebuilding it from prose. The authors want papers to carry an "agent availability" section saying whether a runnable version exists. - [Your Agent's Questions Need a Recipient and a Timer]() — Published 2026-10-30 When an AI agent stops mid-task to ask a human something, it often reads as a failure of autonomy. In one study, letting coding agents ask about bug reports stripped of key details won back most of the lost ground, lifting solve rates from about 55% to 69.4%. But the "human" answering was another model, holding the full report and standing by. Real people are in meetings, or asleep. The design work is in the ask: who gets it, what it says, and what happens while it waits. - [Lebanon's Two Clocks and the Question Calendars Don't Ask]() — Published 2026-10-30 On March 26, 2023, Lebanon woke up on two clocks. The caretaker government had postponed daylight saving three days earlier. Churches and TV channels moved forward anyway, government offices stayed put, and most phones jumped on their own. Middle East Airlines printed every departure an hour earlier, so its planes still left at the same moment. That week made each calendar show what it had been storing: an exact instant, or a time on the wall that politics can move. Software assistants now make that choice without asking anyone. ## Other sections in Issue 46 - [Market Pulse]() - [Foundations]() - [Practitioner's Corner]() - [Echoes]() - [Vision]() ## More - [Publication guide]() - [Issue archive]()