Article URL: https://www.niemanlab.org/2026/10/chatgpt-is-adding-real-cartoonists-signatures-to-fake-new-yorker-cartoons/ Comments URL: https://news.ycombinator.com/item?id=49971846 Points: 53 # Comments: 5
I wanted to build something useful and ended up creating ResilienceLab 0.5.0, an experimental resilience library for .NET 10 with its own API and an MIT license. I developed it with coding agents, reviewing the contracts and checking the results. It includes retries for exceptions and generic results, cooperative timeouts, circuit breakers, fallback, rate limiting, hedging, and an HTTP adapter integrated with IHttpClientFactory. Optional observability uses logs, metrics, and ActivitySource traces. In the catalog API, I composed these layers from outside to inside: Fallback → Circuit breaker →
An SMS provider can deliver and verify a code, but it cannot see enough of your application to decide whether a request is abusive. Short answer: design a secure SMS OTP login flow with per-user, per-IP, and per-device limits in front of every send; give each challenge a short expiry and an attempt ceiling; consume it exactly once; temporarily lock repeated failures; and enforce country policy before spending a send. Provider selection comes after those controls, not before them. For a healthtech marketplace notifying a seller about a new order, keep that notification outside the login challen
This is a submission for the MLH x DEV Writing Challenge Over 80 million people play GeoGuessr. Around 300,000 of them play every single day, staring at real street corners in real cities, for fun. Players of Cities: Skylines have built almost 40 million simulated cities, deciding where roads go and which neighborhoods get a hospital. Now think about how a real city decides where its disaster shelters go. A handful of planners, a set of models, and a public meeting where a good turnout is a room of people.
Article URL: https://globalsolaratlas.info/ Comments URL: https://news.ycombinator.com/item?id=49971782 Points: 7 # Comments: 2
Article URL: https://almostimplemented.github.io/jazz-pianist-style/ Comments URL: https://news.ycombinator.com/item?id=49971726 Points: 9 # Comments: 1
Article URL: https://dbushell.com/2026/10/03/deno-to-node/ Comments URL: https://news.ycombinator.com/item?id=49971719 Points: 16 # Comments: 0
Article URL: https://nivdayan.github.io/dostoevsky.pdf Comments URL: https://news.ycombinator.com/item?id=49971678 Points: 6 # Comments: 0
This is a submission for the Hacktoberfest Open-Source AI Challenge Week 1: Touch Grass Touch Grass Sports AI is a fully offline open-source AI that generates personalized outdoor sports challenges. Its only mission: get you off the screen and into the real world. You pick your fitness level, sport, available time, and location. It replies with a concrete challenge that must be done outside and always includes a real “touch grass” moment. Supports many sports: running, basketball, cycling, yoga, swimming, climbing, parkour, volleyball, tennis, skateboarding, martial arts, hiking, and more.
Hello developers, Today I would like to talk about an error message with a real talent for sending you in the wrong direction: sqlalchemy.exc.TimeoutError: QueuePool limit of size 5 overflow 10 reached, connection timed out, timeout 30.00 (Background on this error at: https://sqlalche.me/e/20/3o7r) If your FastAPI app talks to a database through SQLAlchemy, chances are high you will meet it sooner or later. Not on your machine, of course. It shows up in production, after a few quiet hours, under load. And the reflex is always the same: the pool is too small, so let us make it bigger. pool_size
Hello, my name is Oluwademilade Ala and I am currently studying on computer fundamentals Today's topic will be What is a computer? In our generation or time, there is hardly going to be an individual that hasn't seen a laptop before except maybe you were born in the skies. Today I'll be discussing what actually happens when you pick up your laptop, try to play a game or watch a movie with your girlfriend, or even try to see what number of points Steph curry has on google. Alright, so at the simplest level. We can think of four basic functions and say that a computer is a machine that Recieves
When a media cleanup job outlives the HTTP request that started it, the operational constraint changes the answer. Short answer: use a managed queue for production retries when you want a worker that survives web-process restarts, then record an idempotency key before acknowledging each message. BullMQ is a good fast start for a Node.js team that already runs Redis; SQS is a strong default when AWS operations are already in place; a small managed queue service is attractive when you want less infrastructure to maintain. The mental model is simple. Before: a request calls cleanup, waits, and ow
We read 3,426 public posts from builders whose apps broke at or after launch, and verified 418 recent cases. Database and data loss is 6% of them. Six verified reports in that data describe the same failure: the deploy succeeded, the migration tool says everything is applied, and production disagrees. From the builder's chair it looks like this. API routes return 500 straight after a deploy, and the logs say column "plan_id" does not exist.
Security Audit Report: Reentrancy & Access Control Review: Uniswap V3 Target Protocol: Uniswap V3 (TVL: $1707.6M) Security Audit Report – Reentrancy & Access‑Control Review Protocol: Uniswap V3 (TVL: ≈ $1.71 B across Ethereum & L2s) Audit Scope: Core smart‑contract suite (router, periphery, pool, factory, and related libraries) – focus on reentrancy‑related state‑mutations and privileged‑function access control. Date of Issue: 5 Oct 2026 Prepared By: [Senior DeFi Security Researcher – Confidential] Uniswap V3 is the flagship AMM on Ethereum, featuring concentrated liquidity, multiple fee tiers
How AI Accelerated the Development of GDPR Compliance Software: Building GDPR compliance software has traditionally required extensive legal research, complex engineering, and countless hours spent translating regulatory requirements into practical workflows. Today, AI-assisted development is changing that process dramatically. Over the past year, I've been building GDPR compliance tools using Claude Code, leveraging AI to accelerate coding, documentation, architecture design, testing, and product development. The result is faster iteration, better software quality, and more time spent solving
I use AI every day. I build things with it, experiment with it, write code with it, learn about new things with it, and generally think it's one of the most interesting technologies I've ever had access to. So this is definitely not an "AI is ruining everything" post. But I do think AI is creating a problem that we probably need to talk about more: It has become incredibly easy to avoid thinking. And that cannot be good for us.
We Work Remotely publishes its jobs as RSS, and says so plainly on its RSS page: "Anyone can use the feed, all we ask is that you attribute the links back to We Work Remotely." That makes it one of the few big job boards you can build on without a terms-of-service argument. We built a tool on it at the end of September, and this is what the feed contains once you read all of it, measured on 25 September 2026. There is one feed of the newest jobs, https://weworkremotely.com/remote-jobs.rss, and one per category, listed on the RSS page. The newest-jobs feed held 80 jobs. All twelve together held
On September 8, 2026, Inception Labs announced Mercury 2.5 — which the company describes as the largest diffusion language model ever trained. The headline number: 1,107 tokens per second on widely available NVIDIA GPUs, at quality the company says matches the cost-optimized frontier tier (GPT-5.6 Luna Low, Gemini 3.5 Flash-Lite, Claude Haiku 4.5). It's a 40% intelligence jump over Mercury 2, and it's aimed at one thing the whole industry pretends doesn't exist: the speed ceiling built into how every mainstream LLM writes. Every LLM you've ever used writes like a typewriter. It picks token 1,
We built a tool that takes company names and returns their open jobs: type Databricks, OpenAI, Lyra Health, get their postings as JSON. Behind it is an index of more than 500 companies whose job boards on Greenhouse, Lever and Ashby we had proven belong to them (how, and why guessing gets it wrong half the time, is in an earlier post). The first runs on the real platform found three problems that no unit test had. All three returned HTTP 200, and all three had been sitting in tools that were already public. Input: three companies, 60 jobs.
Every API call you make to an LLM is secretly two jobs glued together. The first reads your entire prompt in one giant matrix multiply — compute-bound, GPUs at full throttle. The second dribbles out tokens one at a time, each step re-reading the entire conversation history from memory — memory-bandwidth-bound, the GPU mostly idling. Run them on the same GPU and each one sabotages the other. The industry's 2026 answer, reached independently by half a dozen labs: stop running them on the same GPU.
