All desks

AI-produced math proofs are arriving faster than people can check and explain them. Who gets to do that work, and how?

Can a history of the British novel rescue forgotten books without shutting out the people who study them?

Four places Cole uses a fast decision model in his AI second brain: email triage, injection screening, news filtering, and deduplication—with a clear boundary for when an LLM still does better.

Florida's sports-betting compact faces a prediction-market challenge involving age limits, college player props and the revenue shared with state and tribal governments.

Michigan, Connecticut and Missouri enforcement is interrupting sports event contracts, although the federal-versus-state jurisdiction fight remains unresolved.

The league backs New Jersey's bid for Supreme Court review of sports-event contracts, citing gaps between federal market oversight and state betting safeguards.

A congressional request for VIP videos and recipient records tests whether loyalty marketing and gambling-harm screening share the same customer data.

Operators contest proposed safeguards for exchange-owned trading desks, exposing a conflict between the peer-to-peer pitch and the economics of affiliated market makers.

Blanche Lincoln's changing position on sports event contracts sits inside a larger fight over whether federal oversight displaces state gambling law.

Invest in accessibility, native CSS, and communication rather than mistaking a fashionable stack for a durable web-development career.

Mateus Fonsek compares three decision-model APIs across Portuguese agent routing, injection screening, and answer grading—with cases, code, and response logs available to inspect.

A case for democracy built on social equality and diplomacy, not technocratic reassurance.

If customer traces are sensitive, seek limited authorized access, recruit cleared domain experts, or carefully redact; synthetic examples are a last resort.

Hamel ranks 28 evaluation practices: inspect real failures, use human-validated product-specific checks, and stop mistaking generic dashboard scores for progress.

Ottawa plans to make a temporary exclusion indefinite, while a government committee also calls for fairer access to mental-health care.

Software posts lose readers through slow openings, assumed knowledge, links that do the explaining, and broken mobile layouts.

A macOS desktop pet routes bounded pose choices to hosted Jev or a native Open-Jev scorer, while voice uses separate models; the code is inspectable but parity tests are opt-in.

An inspectable roguelike puts Jev behind legal choices and a rule fallback; its own paired runs find better style-matching rewards but less variety.

A nineteenth-century history of the Devil tries to reconcile science and religion, but its story of progress has revealing blind spots.

A reproducible decider-2b deployment on Snowflake shows public-benchmark accuracy, GPU scaling, and the idle-capacity bill behind impressive per-decision prices.

Li and Wagle extract typed choices from ordinary LLM token probabilities; their experiments show when narrow fine-tuning helps—and when it damages other decisions.

Paul Gadi reads No Man's Sky through yūgen: travel, isolation, and discovery give its vast universe emotional weight.

Paul Gadi shows how three minigames make familiar genre rules strange again, from commodity-powered flight to chess with cooldowns.

A theory about Claude Code's suggested next prompts meets a direct correction from the team that built them.

Drop in for a free Thanksgiving-themed painting afternoon in Washington Heights on November 7; no experience or registration needed.

A former federal worker's account of what an abrupt firing takes away: stability, a sense of future, and the room to care for family.

Tonight: free vampire cinema in Bushwick and Disney karaoke in LIC; uptown exhibitions, art walks, trivia and a Halloween-TV screening are ahead.

Three skills for turning a spec into reviewed work, making PRs easier to judge, and learning from real agent sessions—plus a breaking glossary filename change.

Wednesday’s AI news is about what you can actually run or measure, and what still needs testing before you trust it. Mistral Large 4 leads because its API preview has independent performance and cost measurements, but promised open weights …

SREGym’s Jev-only diagnosis pipeline passed 80 of 105 judged incident reports; its missed faults expose the limits of evidence selection and single-component answers.

Byren Cheema’s reproducible chess test found fast, cheap, always-legal Jev moves—but almost no wins against opponents trying to win.

A gameplay engineer argues that code is creative game-making labor, and that AI-generated code deserves the same scrutiny as generated art.

Ross Barkan wants little magazines to make room for criticism of new books, not just big arguments built around them.

A party condemns genocide yet rebukes a call for Palestinian equality while overlooking German Green support for Israel's war.

Short practices sometimes match longer ones on stress and mindful awareness. The evidence supports a manageable starting point, with important limits around duration, causation and safety.

Veris put Jev in a Pipecat voice agent's turn gate: fewer caller interruptions across simulated calls, but slower replies and virtually unchanged task success.

Vals AI's two paired evaluations show Jev tying frontier models on filing-based claim checks but ranking last on a small LegalBench slice; a tuned error threshold also missed its target.

Ten GitHub projects for voice, chat, coding, research, images, music, clips, app building, agents, and photo upscaling—with the costs and limitations of going local.

AI can turn an idea into an app quickly. The author asks what happens to the satisfaction of making it.

Inflation policy has a history of making workers pay for problems they did not cause.

The 100-trace target is a starting cue, not a magic sample size; keep inspecting until examples stop teaching you something new.

A Claude Code mod uses Jev to choose effort, models, and context pruning; its cache-switch math and measured-versus-estimated savings are more interesting than its headline savings.

Lee and Berg-Kirkpatrick compare one-pass answer probabilities with Jev across classification, agent monitoring, and triage; task mix and hosting matter more than the headline tie.

A nearly wordless alchemical book reveals how the boundary between experimental science and magic was drawn and redrawn.

Techdirt gathers the early Muse incidents, from Mac access and message privacy to risky agent permissions and a rushed launch.

The Linux maintainer separates useful AI-assisted bug finding from noisy security reports and asks who will actually review, fix, and deploy the patches.

Nightly pickup soccer runs until 2 a.m. at one park in each borough this fall. Adults can join solo or reserve for a team.

Harness compares its production API-inventory pipeline with a single LLM and Jev on a human-checked endpoint sample; the cost story is strong, but misses still matter.

On Cisco's content-safety evaluation, a policy-trained classifier beats zero-shot Jev at strict false-positive budgets; the reference labels and thresholds limit deployment claims.

Josh Bycer on switching between an accessible design concept and the concrete rules needed to implement and tune it.