Monday rebound after the big finale weekend: two free concerts today, a producer meetup in Astoria, plus a full week ahead — Just Sixties at Cunningham Park tomorrow, free comedy Wednesday, and the Miles Davis symphonic show in Harlem on Friday.
Stripe acquires OpenRouter for $7B+; Simon Willison hands-on Qwen 3.8 27B review; Anthropic publishes Claude system prompts; four agent-tooling papers; Daring Fireball on Claude watermarking.
AI code review bots are slow, expensive, and noisy. Swizec argues that custom linters — deterministic, fast, editor-integrated — are the right guardrail for fast-moving teams shipping both human and agent-written code.
Qwen3.8 27B scores 52 on Artificial Analysis — beating all medium models and tying DeepSeek V4 Flash. But its default xhigh reasoning effort makes it spectacularly wasteful out of the box.
Game Code School finished a free 17-part series building a complete roguelike from scratch in C++ and SDL3 — BSP dungeons, shadowcasting FOV, Dijkstra map AI, and a clean architecture, each part shipping a playable milestone.
Frontier models are trading factual knowledge for reasoning skill — and that changes everything about how we deploy them.
A first-time dev's demo flopped and he asked HN how to learn game design. 140 comments later, the answer is consistent: ship many small games, clone mechanics, ask for feedback early, and kill the signup wall.
A gray market of AI inference credit resellers has emerged, offering 30-80% discounts on API tokens from major providers — with tens of millions of dollars in supply.
Anthropic's Frontier Red Team documents how AI agents fail to coordinate — converging on identical strategies, colluding on pricing, escalating turf wars, and struggling with epistemic vigilance — despite individual intelligence.
A solo RPG dev asks how to catch balance regressions when the problem isn't the curve — it's the interaction between crit, attack speed, and enemy resist that nothing catches until weeks later.
Aziz Huq on how partisan gerrymandering, campaign-finance rulings, and dual-class shares are legally locking in concentrated power.
Riley Brown's full content stack on Codex: skills for research, hooks, thumbnails, and diagrams — plus his test-based, outcome-driven way of improving skills.
Storytelling at the Hans Christian Andersen statue in Central Park is a free 70-year tradition running Saturdays through Aug 29 no tickets needed just show up at 11 AM.
Three days of free jazz headlined by the Jazz at Lincoln Center Orchestra, Joshua Redman, and Ravi Coltrane, Aug 28-30 in the neighborhoods where Charlie Parker lived and performed.
Watch a free Revolutionary War reenactment in Prospect Park on Aug 22 — the Battle of Brooklyn's 250th anniversary with period uniformed reenactors, the Maryland 400 commemoration, and historical exhibits.
The canonical resource for learning to design large-scale systems: system design interview prep, architecture tradeoffs, and Anki flashcards. 364k stars.
Finale Sunday: Harlem Day wraps Harlem Week on W 135th St, Puppet Fringe and Battery Dance close out, plus two free concerts, the India Day Parade on Madison, the Cuban Film Fest final day at Anthology, and a free makers market in Greenpoint.
Quiet Sunday: Anthropic details Claude text watermarking and multi-agent failure research; HF State of Open Models summer report; Fields Medalist Gowers on LLM math capabilities.
A devlog on a Python/pygame crime-sandbox roguelike where crime is epistemic — witnesses, disguises, and framing all fall out of what NPCs actually know.
A Godot indie dev self-disclosed his AI-assisted pipeline and still got branded an AI game — but the thread's best replies argue the label tracks perceived cohesion and care, not the toolchain.
A free MIT-licensed desktop app drives a locally installed Codex CLI to generate 2D sprites and animation frames, with an internal sprite harness that keeps characters consistent across frames — the hard problem raw image generation cannot solve.
E. Tammy Kim's weekend essay on presbyopia as the body's unceremonious announcement of middle age — a meditation on aging through the lens of 1.8 billion stiffening eyeballs.
Florence Hazrat's history of punctuation traces the period (2,000 years), question mark (8th c.), exclamation point (14th c.), and why nothing new has stuck since 1600 — with a left political edge the WSJ reviewer found annoying.
The skills that make AI collaboration productive — sharing context, explaining intent, responding to unexpected results — are leadership skills, not programming skills.
Cunningham Park in Queens has free movies, concerts, and archery through late August — no registration required, just show up.
How 1500 LLM-driven submissions turned a CUDA kernel optimization contest into a 232x speedup — a case study in coding agents as research accelerators.
LLMs excel at math not because they reason better, but because they have a vastly larger working memory — the context window removes the key biological bottleneck on human mathematical performance.
Naomi Fry reviews Barbara Kopple's 2002 documentary 'The Hamptons' — a pre-9/11 time capsule of class, celebrity, and summer on Long Island's South Fork
How a family used AI to build free, open-source one- and two-switch games for a quadriplegic nonspeaking brother — with level editors as the core accessibility feature, because no two disabled players have the same needs.
Jan Kandiyali argues that Marx's vision of communism has been misinterpreted as individualistic — it is actually built around mutual self-realization through working for others.
A practical guide to token efficiency, prompt caching, and context management with AI coding agents — from an Anthropic engineer.
Free-music Saturday: Blues BBQ on Pier 76, the NY Guitar Festival finale on the Bryant Park lawn and bluegrass on the UWS tonight — plus a new LES garden-art opening, BOMB's Small Press Flea in Brooklyn, and Harlem Week, Puppet Fringe and Battery Dance all in their final weekend.
Alibaba releases Qwen3.8-27B open weights under Apache 2.0; GLM-5.3 day-2 adds Mythos-5 comparison; 4 fresh agent papers; Anthropic raises misalignment risk estimate.
Lessons in reliability, observability, and failure-mode thinking from building autonomous warehouse robots — and why your next web app should ship with the same rigor.
A solo developer in Johannesburg shares hard-won lessons about scope, feedback loops, and the mindset required to ship a game alone.
A practical GDC 2026 recap filtered for indie teams — what AI tools are actually shippable, what's still demo hype, and what the survey data says about industry sentiment.
Mark Seemann argues that if LLMs write code, the ORM's main justification — reducing SQL boilerplate — collapses. Why not ask the LLM to write fine-tuned parameterized SQL directly?
A self-described AI skeptic argues the gamedev AI question is not black and white: derivative art is a legal and creative liability, while code assistance and reactive personalized narrative are genuinely defensible uses.
Leopold Aschenbrenner blow-up a parable for the AI industry broader failure of intellectual humility — overconfident experts overreaching beyond their domain.
Z.ai shows that scaling RL post-training on diverse, long-horizon coding environments produces rapid gains — and an unexpected side effect: emergent cyber exploitation capability.
A Reddit dev demonstrates single-prompt game generation via extended AI thinking — producing a playable Cuphead-inspired boss-rush for ~$30 with zero pre-existing assets.
Anna Russell on the double bind of reproductive timing: AOC goes public with egg-freezing, and the real failure is structural, not personal.
A 15th-century Zoroastrian divination text that tells you what it means to see a snake on each day of the calendar — remarkable because Zoroastrians traditionally consider snakes evil creatures of Ahriman, yet half the omens are auspicious.
The Tin Building by Jean-Georges has closed after four years. In its place: Balloon Museum, an immersive inflatable art experience opening this summer at Pier 17.
Amazon has spent $5M+ fighting a New York City bill that would force it to directly employ delivery drivers instead of subcontracting through middlemen — a fight with national implications for labor organizing.
A defense of programming as a genuinely hard craft, and a rejection of the cliché that code was never the hard part — without romanticizing it into unautomatable art either.
Eight giant illuminated seesaws that light up and play music as you ride. Free, no reservations, daily 9am-9pm at Herald Square through August 21.
AI is already embedded across game production workflows. Refusing to engage with it is becoming a professional risk — the useful question is where the tool helps and where it fails, not whether to reject it outright.
Playtesters told a zero-budget solo dev that AI art would turn players away. The thread's consensus: detectable slop is the turn-off and a hard ~15% of players are unreachable — but cohesive, deliberate work passes.
Roose and Newton fact-check Zuckerberg's 6,500-word AI manifesto — policy wishlist dressed as optimism, with a dragon-to-everyone critique — then Pangram CEO Max Spero explains how AI text detection finally got good, and the hosts debut Running the Numbers with a Riemann hypothesis side problem solved via positive affirmation.