Sharon Horgan, Autofiction Auteur — The New Yorker

Horgan’s new HBO comedy Youth turns the sandwich generation and midlife dating into half-hour comedy — and she explains why her shows keep aging with her.

Published:  · Added:  · 3 min

MCP Is Not Just Another API Standard — Balaji Venkatasubramaniyar

MCP standardizes how agents find and call capabilities, but the hard integration decisions move into model behavior at runtime—and demand better descriptions, testing, and controls.

Published:  · Added:  · 2 min

Will TypeSafe’s Jev Change How We Build AI Applications? — Laurie Voss

Laurie Voss makes the architectural case for cheap typed decisions, calibrated probabilities, and a hybrid Jev-plus-LLM evaluation loop—while separating launch claims from early tests.

Published:  · Added:  · 3 min

Explaining to Business People Why Building Software Is Still Hard — Anton Zaides

Anton Zaides on why software resists estimation: the house-renovation analogy for infrastructure, refactoring, and root causes — and why day-two slowdowns are normal.

Published:  · Added:  · 2 min

SAML: A Fractal of Bad Design — Matt Schwager

Trail of Bits argues SAML has earned retirement: canonicalization, enveloped signatures, and kitchen-sink spec design keep producing auth bypasses 25 years in.

Published:  · Added:  · 3 min

Zach Cregger Made a Resident Evil Movie With Bite — Jacobin

Eileen Jones on a reboot made by a director who has never seen the franchise: fast, funny, scary, and a black comedy about a culture that will not admit the catastrophe is real.

Published:  · Added:  · 2 min

Will OpenAI Eat Jev's Lunch? — John Berryman

Jev returns calibrated probabilities instead of sentences. OpenAI has been using LLMs as classifiers for years, so if the moat is only in the training data, the trick can be folded straight into its own models.

Published:  · Added:  · 7 min

The Freedom of the Weasel — n+1

Grace Glass on Annie Dillard: the writer who taught fifty years of American essayists how to notice, and whose god's-eye view could not see the occupation in front of her.

Published:  · Added:  · 2 min

I Asked Meta's Muse for Its Filesystem and It Sent Me 6.8 GB — Pete at mouse.dev

One conversation and a Google Drive destination were enough to pull 6.8 GB of the agent's own runtime — internal docs, memory files, 68 skills, and SSH keys. Meta marked the report Not Applicable.

Published:  · Added:  · 6 min

If AI Coding Is Lowering Your Code Quality, You're Not Managing Quality Right — Iouri Khramtsov

Seven defensive layers — spec review, near-universal unit tests, E2E gates, targeted AI quality passes, dual AI review, production monitoring — that let one team roughly double output without giving up reliability.

Published:  · Added:  · 2 min

Engineering Memory — Gregory Gundersen

Gregory Gundersen memorized 100 digits of pi in a single sitting, not through talent but by turning recall into a system: a person-action-object index, a memory palace, and errors that stay local.

Added:  · 2 min

Village Halloween Parade 2026: The Free Parade You Can Actually March In

The 53rd Village Halloween Parade is Saturday, Oct. 31, 2026, stepping off at 7 p.m. up Sixth Avenue. Anyone in costume can march for free — lineup entry at Canal Street between 6:30 and 9 p.m.

Published:  · Added:  · 2 min

Should Routine AI-Assisted Code Require a Disclosure Label?

A Unity Asset Store publisher asks whether AI-written code needs the same AI label as AI art. The thread lands on the real fault line: disclosure works when the label carries scope and can be checked — and for code, it can't be checked.

Added:  · 3 min

Claude Opus 5.5 Is Here - Is Claude Finally Back? (5 Use Cases Tested)

Peter Yang stress-tests the new Opus on five tasks: a Golden Gate Bridge 3D flyover in Blender, a playable Soaring Over the World ride artifact, stroke-by-stroke impressionist painting, Claude Design inside Claude Code, and frame-by-frame video editing - plus the personality prompt that Opus 5 failed.

Published:  · Added:  · 4 min

Frontier AI on Your Own Hardware — Tim Dettmers

A CMU professor argues that cheap-to-run local models and autonomous agents are about to hand the next decade of AI research to university labs — and the 94-comment thread is mostly about whether the claims add up.

Published:  · Added:  · 6 min

Even Ed Sheeran's Apology Took Both Sides — The New Yorker

Ed Sheeran let Macklemore be dropped from his tour for saying Free Palestine, then stood alone in Philadelphia and apologized twice — once for apologizing.

Published:  · Added:  · 3 min

Can gzip be a language model? — Nathan Barry

A beam search over compressed output turns the archiver that ships with your operating system into a text generator — plus what 122 HN commenters had to say about where the trick holds and where it falls apart.

Added:  · 5 min

The Last Revolutionary Filmmaker — Jacobin

Masao Adachi spent decades in Lebanon alongside the Palestinian resistance, was jailed and stripped of a passport — and at 87 is still making cinema that refuses to split art from revolt.

Published:  · Added:  · 3 min

The Economics of Open-Weight Inference — Ornn Data

A GPU-rental data firm argues that the standard "new generation kills the old one" depreciation model is wrong — because open weights give older silicon a workload it can still win.

Published:  · Added:  · 4 min

Kalshi, Tribes Appellate Ruling Could Grab SCOTUS Attention — Michael McCann

A second Ninth Circuit decision, this one on tribal lands, gives the Supreme Court a different vehicle for deciding whether event contracts are sports bets — and puts IGRA, UIGEA and tribal sovereignty into the same case as the Commodity Exchange Act.

Published:  · Added:  · 4 min

How Linear Made CI Faster While Its Test Suite Quadrupled — Mufeez Amjad

CI cost is mostly fixed cost per job — runner boot, checkout, install. Linear cut PR wait below five minutes and halved runner time per test by attacking setup, gates and sharding order, not the tests.

Published:  · Added:  · 2 min

Why Does an npm Math Library Need an Encrypted Loader? — SafeDep

mathmain@1.0.0 hides a remote access implant behind AES-GCM: the decryption password is the LU lower factor of whatever matrix you pass to the solver, so nothing sensitive is greppable in the published code.

Published:  · Added:  · 2 min

AI Has No Wisdom and Neither Will You — Alexandru Nedelcu

Alexandru Nedelcu's case against treating code as disposable output: maintainability has no reward signal a model can learn from, and the developers who stop reading code stop learning from their own mistakes.

Published:  · Added:  · 8 min

Free Forest Fitness at Fort Tryon Park - Stairs, Tree ID, and Forest Bathing

Fitness at the Fort is a free, year-round workout in Fort Tryon Park: Tuesdays and Thursdays 7:30-8:30 a.m., stair climbs, stretches, tree identification and forest bathing. Meet at Margaret Corbin Circle.

Added:  · 2 min

Works in Public: Two Free Sculptures Land on Riverside Park South's Locomotive Lawn

The Art Students League and NYC Parks open Works in Public on Monday, Oct 5, 3:45-4:30 p.m. at Locomotive Lawn in Riverside Park South - sculptures by Alicia Ehni and Robert Franca, free, RSVP on Eventbrite.

Added:  · 2 min

Balancing a Game the Right Way: Make Stats Designer-Facing

Rename your balancing variables around what a designer can picture — wait time instead of depletion rate, heat radius instead of heat strength — and let the code do the conversion.

Published:  · Added:  · 2 min

I Said No and Apple Said Yes — David Bushell

A web developer's log of Apple Intelligence: the opt-out he used in 2025 is gone in macOS 27, plus 22 GB of models he never asked for and 271 HN commenters on what to do about it.

Published:  · Added:  · 6 min

Imagining an Idle Countess: George Wightwick's The Palace of Architecture — The Public Domain Review

In 1840 an architect wrote a world history of architecture for idle women — and the profession's response shows exactly what it thought it was protecting.

Added:  · 3 min

Spymarks, Not Watermarks — Brandon Thomas

A new name for hidden tracking signals embedded in AI-generated images, audio and text, and the argument that the word watermark hides what SynthID-class systems actually are.

Published:  · Added:  · 6 min

Zack Polanski's Eco-Populism Could Do With More Socialism — Jacobin

A year after Polanski took over the Greens, membership has nearly quadrupled — but the party's electoral aims contradict each other, and its own version of socialism stops at the slogan.

Published:  · Added:  · 2 min

Kalshi asks the CFTC to allow margin trading on event contracts — Davis Giangiulio

Kalshi's clearinghouse asked the CFTC to approve margin for event contracts — sports stay fully collateralized, margined contracts are described as institutional-only, and the filing's own text leaves the retail FCM channel open, with the methodology redacted.

Published:  · Added:  · 5 min

NYC Digest - 2026-09-22

Normal Tuesday, 8 items and all 8 free: Laraaji's equinox set and other free garden music in the LES tonight, the free SummerStage show in Central Park, a new free SoHo exhibition opening today, and the Studio Museum's first uptown window tomorrow.

Added:  · 4 min

AI News - 2026-09-22

Tuesday runs on two tracks: open weights and agent security. The lead is Xiaomi’s MiMo-V2.6 — a 1T-parameter MoE under an MIT license that Artificial Analysis scores at 46, the top of 114 measured open-weight models and level with Grok 4.7, …

Added:  · 9 min

Climate Anxiety and the Letter That Came Back

A 14-year-old with climate dread wrote to imprisoned climate activists instead of waiting to feel better. The Guardian's new documentary follows what the replies did — and ends with his letter surfacing in a courtroom.

Published:  · Added:  · 2 min

Sean Chen’s Jev Judgment Race: Fast Answers, Small Test Set

A hands-on walkthrough compares Jev with Opus, Haiku, and GPT-5.4 Mini on bounded decisions; the speed is striking, but the labels and tiny sample do not establish general accuracy.

Published:  · Added:  · 2 min

I Am Often Wrong — Boris Cherny

Boris Cherny's six-step loop for problems and products, and the two steps where plans actually break: defining the problem, and defining a clear and simple approach to it.

Published:  · Added:  · 2 min

Unpatched libheif and an SSO Flaw: Chaining an Image Upload to Account Takeover — Hacktron AI

A libheif fix that was never marked as a security fix never got backported, so Debian still shipped the bug. Hacktron chained that decoder flaw through a forum image upload into an SSO identity failure, and from there into connected developer accounts.

Published:  · Added:  · 3 min

Design Lesson 101 — Metal Gear's Inverse Difficulty Curve

A 2008 column by Raven designer Manveer Heir argues Metal Gear front-loads its difficulty — unarmed and ration-starved at the start — and swaps which skill the game tests as you get stronger.

Published:  · Added:  · 2 min

It's My Park: 46 Free Volunteer Workdays in NYC Parks This Fall

NYC Parks' It's My Park season runs into December with free, walk-on workdays — trail maintenance in Van Cortlandt, bulb planting in Owl's Head, a Little Free Library ribbon-cutting in Brookville Park and a five-site Saturday in late September.

Added:  · 3 min

NYC Parks Fall Festivals 2026: Ten Free Events Through Halloween

NYC Parks' fall festival calendar runs from Oct. 3 through Halloween and every listing is free — a Lenape pow wow, a Rockaway pumpkin patch, harvest parties and Bronx Halloween bashes. Dates, times and locations.

Added:  · 2 min

The Curious Power of Punctuation — The New Yorker

Louis Menand reviews two new books on punctuation and verbs, and finds in their love of reading as a bodily experience an elegy for a way of reading that machine-generated prose is ending.

Published:  · Added:  · 2 min

Off with Their Heads: Illustrations of Blemmyes (ca. 1175–1724) — The Public Domain Review

Five centuries of illustrations of the Blemmyes — the headless humanoids of ancient geography whose eyes and mouths sat in their chests — and what their long afterlife says about how travel writing turns strangers into monsters.

Added:  · 2 min

I Don't Want to Read What You Didn't Write — Colin Breck

Design docs, PR notes and meeting summaries now arrive generated instead of written. A systems engineer explains the context gap that makes them unreadable — and the narrow set of jobs where AI genuinely improved his writing.

Published:  · Added:  · 7 min

Media Capture, End to End: Three Pieces on How American News Got Bought

Pickard's model, Sullivan's mechanism, Poynter's architecture — a February–March 2026 reading path through the same story, plus the timeline that ties them together.

Added:  · 5 min

The Ellisons' Empire: Media Consolidation, Narrative Control, and the Threat to Democracy — Nonprofit Quarterly

A Pentagon press conference, a $111 billion merger, and the clearest case yet that narrative control can simply be bought.

Published:  · Added:  · 4 min

CBS News Is Convulsing as Larry Ellison Tries to Please Trump — The Guardian

Cooper leaving 60 Minutes, Colbert taking his interview to YouTube, a producer's farewell note about editorial independence — and a merger that needs the FCC's blessing.

Published:  · Added:  · 3 min

AI Sovereignty Needs the Whole Open Stack — Nicole Butterfield

Open model weights alone do not give countries or institutions control over their AI. The dependencies run through chips, cloud, agent harnesses, data systems, and observability.

Published:  · Added:  · 1 min

GOP Senator Wants Probe Into Trump Jr.'s Prediction Market Ties — Corey Sharp

A senator's call for subpoenas puts a direct conflict question to prediction markets whose national reach depends on federal regulators.

Added:  · 2 min

The American Media Polycrisis: Cascading Layers of Capture — LPE Project

Not one collapse but stacked layers of capture — capitalist, oligarchic, authoritarian — and why the usual fixes only touch the middle of the stack.

Published:  · Added:  · 3 min

Alternatives to MinIO for Single-Node Local S3 — Robin Moffatt

MinIO was abandoned by its vendor in late 2025, breaking local S3 emulation in demos and build pipelines. Robin Moffatt tests the replacements with a working DuckDB and Iceberg smoke test, then scores each one on config, licence, and bus factor.

Published:  · Added:  · 2 min