When analytics on Postgres slows down, most teams add a second database. Then come the pipelines, the sync jobs, and a copy of your data that's always a little behind.

TimescaleDB takes a different approach: extend Postgres instead of splitting away from it. Hypertables partition your data automatically as volume grows. Hypercore compression cuts storage up to 95%. Continuous aggregates keep dashboards live without re-querying everything.

CERN runs Postgres this way for sensor data from the Large Hadron Collider.

No split architecture, no pipeline lag, no new query language to learn. Same SQL, same drivers, same tools.

Start on Tiger Cloud and get $1000 in credits.

Get $1000 Credit

The White House's AI Task Force Tells Labs That Reporting Incidents And Fixing The Harm Is A National-Security Duty: "Delayed Notification" Will "Not Be Tolerated": Late Friday, Oct. 9, after this newsletter's last send, the Super Intelligence Force told AI companies to report incidents involving their models and fix the resulting harm, framing both as obligations rather than voluntary steps: "Delayed notification, inadequate corrective action, and a failure to take responsibility will not be tolerated." The statement, reported by Bloomberg, followed Anthropic's disclosure that one of its models submitted an invented tip to a police department's homicide form during testing (Safety, below). It names no rule, deadline or penalty, and no order or Federal Register notice has followed, so the duty rests on the force's say-so for now. It fits the administration's stated preference for liability over new regulation, Treasury Secretary Bessent's line to Congress that "the creators are liable for what they build and generate," while the Hawley-Murphy and Trahan liability proposals have not moved. Link

Washington Lets U.S. Labs Send Unreleased Models To Britain's AI Security Institute Again, Ending A 17-Day Freeze: The Telegraph reported Sunday, Oct. 11, that the administration reversed the Sept. 24 request from the Office of the National Cyber Director that labs hold new models back from UK testers until a U.S. national-security review was done; pre-release access for Anthropic, OpenAI and Google DeepMind models is restored. During the freeze Anthropic released Claude Mythos 5.1 through trusted-access programs to a set of U.S. organizations only, without the institute's review. UK AI minister Kanishka Narayan and other ministers lobbied for the reversal. There is no White House statement, the outcome of the review is unreported, and nothing yet confirms Mythos 5.1 goes to the UK now. Allied pre-release testing, the backstop the voluntary accord leans on, turns out to rest on a request that can be withdrawn without notice. Link

IMF Chief Calls AI A "Slow-Moving Tsunami" For Jobs And Tax Bases; The World Bank's Jobs Council Wants Crisis-Era Youth Support Made Permanent: On an AI-and-jobs panel at the IMF and World Bank annual meetings in Bangkok on Monday, Oct. 12, Kristalina Georgieva said about 40 percent of jobs worldwide, and about 60 percent in advanced economies, will be augmented, downgraded or eliminated, that AI moves income from labor, the traditional tax base, toward capital, and that growth in low-paid construction work is "not a story to celebrate"; she made no specific tax recommendation. Singapore President Tharman Shanmugaratnam, co-chair of the World Bank's Jobs Council, urged governments to make crisis-era employment support a permanent tool for young workers, with hiring incentives for graduates. For legislators drafting AI workforce bills for January, it shifts the argument from job counts to the payroll-tax and income-tax base that funds unemployment insurance. Link

Arizona's Attorney General Asks Regulators To Investigate Whether APS Shifted Nearly $600 Million Of Large-Load Costs Onto Other Customers Before Any Rate Increase: In a letter released Friday, Oct. 9, Kris Mayes asked the Corporation Commission to investigate promptly and publicly whether APS's Power Supply Adjuster moved costs from a handful of large energy users, data centers among them, onto everyone else, citing consumer advocate Abhay Padgaonkar's analysis, filed Oct. 2, that $587 million was allocated that way from 2023 through 2025. She says the commission can revisit past decisions, order APS to produce billing and adjuster records and require a full cost-of-service study, and should do so before it rules on APS's $608.7 million rate request: "The Commission should not just take APS' word for it." APS calls the analysis misleading, and the commission's executive director has also disputed it. KJZZ reported the letter Saturday. It moves the ratepayer fight from tariffs for future load to money already collected through a fuel clause. Link

On Local Calendars This Week: Cleveland Moves To Extend Its Pause, Charlotte Hears An Extension Tonight, Shelby County Weighs Its Own, Colorado Springs Petitioners Need 24,861 Signatures: Cleveland Utilities Chair Brian Kazy plans to introduce legislation today for a 90-day extension of the pause on standalone data centers council passed 14-1 on July 15, the one extension the ordinance allows, "to assure that we get this thing right"; the city lists the current pause as running through Oct. 20. Charlotte's council holds its public hearing tonight on carrying the moratorium that expires Nov. 5 to Oct. 11, 2027, with a vote reported for Oct. 26. In Shelby County, Commissioner Henri Brooks's ordinance for a pause of up to 12 months in unincorporated areas, with a development agreement on infrastructure, aquifer protection, energy and noise, left committee without a recommendation Oct. 7 and needs three readings, though a county lawyer has said the commission lacks the authority; Memphis's own vote stays Oct. 20. In Colorado Springs, organizers are circulating a four-year moratorium petition the clerk approved Sept. 30 and need 24,861 valid signatures by Dec. 7 to reach the April 6, 2027, ballot. Link

Anthropic Says Its Models Exploited Server Flaws, Submitted Real Government Forms And A Fake Homicide Tip During Testing, And Cuts Live Internet From All Internal Evaluations: In a post Friday afternoon, Oct. 9, Anthropic described four kinds of unintended actions on live systems found in a transcript review it began in July: exploiting a software flaw to run commands on a university-hosted analysis tool; submitting forms it was told not to, including a real government form after a practice version failed to load; reaching gated data with access tokens taken from a local government map and a state agency's public dashboard, the latter to avoid a data fee; and routing around fetch limits through free URL shorteners. Claude Haiku 4.5 submitted an invented tip on an unsolved homicide to a police department's form; it was flagged as spam, Anthropic told the department Oct. 8, and the Philadelphia Police Department disclosed it Oct. 9. Some cases involved federal, state and local sites. Anthropic briefed the White House, notified each agency, moved evaluations offline, extended its live-internet cutoff to all internal evaluations and built detection tooling it says "blocked all of them" in testing, while saying the cases "do not change our overall view of Claude's alignment." It is the fact pattern behind the task force's reporting demand above. Link

Nadella Says Every Frontier Model Should Be Treated As Already Compromised, With A Human "Emergency Brake" Kept Outside The Model: In a long post on X on Saturday, Oct. 10, Microsoft's CEO wrote that "we must assume a model is compromised and contain it from the start," that closed and open-weight frontier models should be handled as insider risks, that an authorized human must be able to pause or stop a model mid-task, and that access and action controls belong outside the model, with meaningful actions leaving "tamper-proof human readable evidence." "We need to separate the supply of intelligence from the authority over it," he wrote; transparent reasoning is "non-negotiable" but not sufficient, and his seven principles include no dependence on a single model and incident disclosure shared across the industry. It came the day after the task force's statement and reads as a vendor's specification for what a statutory kill-switch requirement, such as the bipartisan Lieu-Moran bill, would ask of deployers. Link

Britain Will Study How Serious AI Incidents And Near Misses Should Be Reported And Weigh Loss-Of-Control Amendments To Its Cyber Bill: The package announced Monday, Oct. 12, by the Cabinet Office, the AI Security Institute, the National Cyber Security Centre and DCMS, alongside a speech by AI minister Kanishka Narayan at Imperial College, will examine how serious AI incidents and near misses should be identified and reported, which the Financial Times reports ministers are considering making compulsory, and whether the Cyber Security and Resilience Bill should be strengthened to protect essential services from AI risks including loss of control. It extends the frontier-AI cyber defense pilots, which the government says have found more than 1,000 vulnerabilities in government systems. "I want Britain to become one of the hardest countries in the world to attack using artificial intelligence," Narayan said, pledging to legislate "wherever necessary." The government rejected a kill-switch amendment to the same bill in September; the Commons Business and Trade Committee questions the labs on pre-release testing Tuesday. Link