TrendPulse

OpenAI is launching two new models, which it says are cut from the same cloth as Astra.

This academy for promising high school grads is somewhere between a trade school, Y Combinator, and Peter Thiel's Fellowship Program.

The theft of agents' personal information could present a major counterintelligence threat, where agents and their families are extorted into cooperating with a foreign government.

Meta says Muse was built from scratch, but acknowledges the AI assistant was "heavily inspired" by OpenClaw — down to some of its workspace filenames and content.

Apple may be developing a new fitness tracker as part of its new generation of hardware devices.

Qualcomm said that its new top chip can run 30B mixture-of-expert model locally

I've been building this for two weeks and want honest technical feedback before I invest more time. What it does Bhive is a multi-agent framework in TypeScript. Agents ("Bees") can delegate tasks to each other, call tools, and it works with any LLM provider — Gemini, OpenAI, Anthropic, or local Ollama — through the same code. Architecture decisions I'm unsure about Provider abstraction via an LLMAdapter interface, with a ToolCallingParser that normalizes too...

I recently completed a system design and performance optimization cycle for CrimeLens , my full-stack crime reporting, mapping, and analytics platform. CrimeLens connects citizens, public users, police officers, and administrators through different workflows. Citizens can submit crime reports, police and administrators can review and manage them, and approved crime data becomes available through maps, statistics, trends, and geospatial searches. The system is built as a PERN modular mono...

PR #466 — fix(citations): enforce document order for [[PAGE_BREAK]] segments Repo: open-legal-products/mike — 4.3k ⭐, open-source legal AI platform for document review, drafting, and legal research. Built with Next.js, Express, Supabase, and Cloudflare R2. The bug I fixed: Citations were verified segment by segment with no position check. So B[[PAGE_BREAK]]A earned verified: true even when A preceded B in the document — wrong order, wrong result. I enforced document ord...

Your hero section is height: 100vh . Big background image, headline centered, a CTA button pinned near the bottom. You check it in the browser, check it in the iOS simulator, ship it. Then the support tickets come in: "the sign-up button is cut off." You open it on your own phone to confirm — and it looks fine. Scroll down an inch, though, and there's a sliver of button peeking out from under where the fold should be. Guess what's different between your simulator and a real phone in so...

Your team wrote middleware.ts carefully — Edge-safe imports only, jose instead of jsonwebtoken , no direct database calls — because that's what Edge middleware demanded. Then you upgraded to Next.js 16, skimmed the release notes, and moved on. Nothing broke. Which is exactly the problem: middleware.ts still runs, but it's now the deprecated way to do the one job every non-trivial app needs — checking a request before a single line of your app runs. The framework renamed the file, moved ...

Design wants a country picker. Nothing fancy — a dropdown, a small flag icon next to each name. You reach for <select> , type the first <option> , and hit the wall every frontend dev hits eventually: an <option> can only ever show text. No <img> , no icon span, nothing. The browser renders whatever markup you put inside it as a plain string. So you do what everyone does. You build a fake one. The dropdown you build instead A <div role="listbox"...

You build a useCounter() composable, drop <Counter /> on the page twice, and click the first button. Both counters go up. You didn't copy-paste a bug. You wrote twelve lines of completely ordinary-looking Composition API code, and Vue executed every one of them correctly. The surprise isn't a Vue bug — it's a gap in your mental model of what a composable actually is , and that gap is exactly where a much more expensive version of this same mistake lives: on a server, where "the oth...

I've been using AI to write more code, which has also meant writing more tests—including screenshot tests to catch unexpected UI changes. That leaves me reviewing two things: the code in GitHub and the visual diff in a Playwright report. I kept switching between them to answer a small question: did that button move, or did its padding change by a couple of pixels? GitHub's side-by-side, swipe, and opacity views help compare images. What I wanted was explicit pixel highlighting, especially ...

I spent forty-eight hours arguing with an API that was not actually slow on the wire. My Python probe kept printing extra delay, while a single curl call still looked almost instant. Have you ever trusted a client-side timer and then shipped a performance scare that never existed upstream? The disagreement showed up during a quiet check, not during an incident bridge or a customer ticket. I had a tiny loop hitting the same health URL and printing elapsed time after every response. curl made t...

goto is absolutely underrated. it isn't inherently bad. like any control-flow tool, it depends on how exactly you'd use it. proper usage: gotoif (error!=0) {do_cleanup} // complete_operation label do_cleanup sys.unlock_file(descriptor) sys.free(buffer) return this is pretty simple: if something goes wrong, go clean up. messed-up usage: @label b goto a @label c goto b @label a goto c // a forever do-nothing loop this is technically valid, but...

Google has expanded its AI-enabled Search Live experience with Gemini 3.8 Live , a new model designed for voice-based dialogue with near real-time reasoning, visual context and language switching. The rollout makes Search Live a more conversational part of Google Search, rather than a tool built only around typed queries and static result pages. According to Google's official Gemini 3.8 Live announcement , the company is also introducing Gemini 3.8 Live Extended Thinking, a version intended...

Top 5 MCP Gateways in 2026 Emmanuel Mumba Emmanuel Mumba Emmanuel Mumba ...

Comments

Comments

Comments

Comments

Comments

Comments

Comments

Hey HN, Guanming here, cofounder of General Instinct. We just released InstinctFlash, a high-performance serving framework for robotics models. It’s licensed under AGPL-3.0. On Jetson Thor, we see speedups about 1.2x to 7.9x from runtime optimizations alone and up to 33.78x for LingBot-VA when we combine those runtime optimizations with a distilled few-step diffusion scheduler, going from the original 25 visual / 50 action steps to 2 / 4 steps. Across 50 Robotwin2.0 tasks, we evaluated 1,153 epi...

Article URL: https://minimaxir.com/2026/09/agentic-iteration/ Comments URL: https://news.ycombinator.com/item?id=49803085 Points: 78 # Comments: 38

Article URL: https://allpoetry.com/16-bit-Intel-8088-chip Comments URL: https://news.ycombinator.com/item?id=49803635 Points: 82 # Comments: 12

Hi HN, I built ai·rete·rag because I kept seeing teams put an LLM in charge of decisions that need to be auditable (lending, fraud, clinical triage), then bolt on "guardrails" after the fact. It runs the two in series instead: 1. A pure-Python Rete engine evaluates YAML rules against your facts. The verdict comes only from here. Same facts, same verdict, every time, with salience-based conflict resolution. 2. RAG retrieves passages from your own policy documents, and an LLM writes a plain-Englis...

Article URL: https://www.anthropic.com/claude-opus-5-5 Comments URL: https://news.ycombinator.com/item?id=49803892 Points: 764 # Comments: 607