AI News Today

Article URL: https://www.ft.com/content/c5d52f72-71ef-40bc-bad3-61afdba8b378 Comments URL: https://news.ycombinator.com/item?id=48704836 Points: 3 # Comments: 0

<p>Most CLAUDE.md files are 500-line monoliths. When you switch LLMs, you rewrite everything. After the third rewrite, I built a three-layer architecture that makes model swaps trivial.</p> <h2> The Problem </h2> <p>I run DeepSeek V4 Pro as my daily driver for Claude Code. But sometimes I need Claude Opus for complex reasoning, or Sonnet for fast iterations.</p> <p>Every time I swapped, I rewrote my entire CLAUDE.md. DeepSeek needs tighter tool-call discipline. Claude Opus needs less output spli

Article URL: https://thehill.com/opinion/technology/5942757-ai-demands-new-social-norms/ Comments URL: https://news.ycombinator.com/item?id=48704759 Points: 2 # Comments: 1

Article URL: https://www.cnbc.com/2026/06/26/oracle-stock-ends-worst-week-since-2001-as-investors-dwell-on-finances.html Comments URL: https://news.ycombinator.com/item?id=48704720 Points: 4 # Comments: 1

<p><em>The Model Context Protocol ecosystem exploded to nearly 20,000 servers. Most are noise. I installed, wired up, and stress-tested 100 of them — mostly inside Claude Code — to find the handful that actually earn a permanent slot in your config. Here are the 12 that survived, the ones I uninstalled, and the uncomfortable 2026 truth nobody selling you MCP servers wants to admit.</em></p> <h2> Why I Went Down This Rabbit Hole </h2> <p>When Anthropic open-sourced the <strong>Model Context Proto

<pre style='white-space:pre-wrap;width:81ex'>test: promote OpenAI HTTP QA coverage (#97369)</pre>

<p>There is a workflow inside your company that everyone quietly works around.</p> <p>Nobody officially owns fixing it.</p> <p>Everyone knows it is painful.</p> <p>New hires learn it through screenshots, Slack threads, and “ask Priya, she knows how this works.”</p> <p>A spreadsheet sits in the middle of it.</p> <p>A manager checks it manually every Friday.</p> <p>A customer probably feels the delay, even if they never see the process.</p> <p>That workflow is not just annoying.</p> <p>It is a tax

Dilip Asbe said that newer UPI apps could be more competitive with a viable commercial model

Article URL: https://warmapper.substack.com/p/how-to-spot-ai-maps Comments URL: https://news.ycombinator.com/item?id=48704490 Points: 3 # Comments: 0

Article URL: https://github.com/Tylersuard/Synapse_neural_net_training_game/tree/main Comments URL: https://news.ycombinator.com/item?id=48704463 Points: 2 # Comments: 1

Article URL: https://github.com/Adirdabush1/cerberus Comments URL: https://news.ycombinator.com/item?id=48704458 Points: 3 # Comments: 0

Article URL: https://github.com/suhasbhairav/ai-chief-of-staff Comments URL: https://news.ycombinator.com/item?id=48704381 Points: 3 # Comments: 0

Article URL: https://code.intellios.ai/photo/ Comments URL: https://news.ycombinator.com/item?id=48704343 Points: 2 # Comments: 1

Article URL: https://github.com/tam159/next-role Comments URL: https://news.ycombinator.com/item?id=48704340 Points: 2 # Comments: 0

Often times I think, why aren't we using more and more agents to often improve themselves. Surely they're smarter than most of us. What do you think is stopping them? A bunch of companies have raised insane amount of capital as well for this, but it's still not productised/generally accessible? Sort of baffles me. Comments URL: https://news.ycombinator.com/item?id=48704307 Points: 2 # Comments: 2

Article URL: https://github.com/xbtlin/ai-berkshire Comments URL: https://news.ycombinator.com/item?id=48704305 Points: 2 # Comments: 0

Article URL: https://wordnamefire.com/ Comments URL: https://news.ycombinator.com/item?id=48704265 Points: 2 # Comments: 0

<p> Stop re-explaining your project to AI coding agents </p> <p> <a href="https://www.producthunt.com/products/pmb-local-first-memory-for-ai?utm_campaign=producthunt-atom-posts-feed&utm_medium=rss-feed&utm_source=producthunt-atom-posts-feed">Discussion</a> | <a href="https://www.producthunt.com/r/p/1182762?app_id=339">Link</a> </p>

We have published the real-time metrics for Exposed DB, Exposed Credentials, Exposed KEV Exposures, Subdomain Takeover: https://echelongraph.io/exposed-databases https://echelongraph.io/leaked-credentials https://echelongraph.io/exposed-ai-keys https://echelongraph.io/kev-exposure https://echelongraph.io/subdomain-takeover Numbers are in real-time based on our scanners. Comments URL: https://news.ycombinator.com/item?id=48704216 Points: 2 # Comments: 0

Article URL: https://mainichi.jp/english/articles/20260627/p2g/00m/0na/029000c Comments URL: https://news.ycombinator.com/item?id=48704142 Points: 2 # Comments: 0

Testing LLM apps and agent frameworks against real APIs is expensive, rate-limited, slow, and non-reproducible. LLMSim is a Rust simulator for the OpenAI Chat Completions, OpenResponses and in future other protocols. LLMSim could be used in two forms, one is the server, with high concurrency for load/stress testing (~40k req/s with p99 ≈ 5ms on 4 vCPUs, scaling with cores); and second it embeds directly as a crate in your tests - no separate process, no network, deterministic. It focuses on traf

Article URL: https://twitter.com/zicohacks/status/2070401037018788301 Comments URL: https://news.ycombinator.com/item?id=48704059 Points: 4 # Comments: 0

Article URL: https://mollykinder2.substack.com/p/we-cant-retrain-our-way-out-of-ais Comments URL: https://news.ycombinator.com/item?id=48703971 Points: 3 # Comments: 0

Comments

<table> <tr><td> <a href="https://www.reddit.com/r/MachineLearning/comments/1uhlvjv/nagatranslate_building_a_translation_and_voice/"> <img src="https://preview.redd.it/bu6xsk4hvx9h1.jpg?width=140&height=63&auto=webp&s=0a4c589616d351d10d735940874706494e48d408" alt="NagaTranslate: Building a translation and voice pipeline for low-resource Nagaland creoles (Whisper, VITS, LLMs) [P]" title="NagaTranslate: Building a translation and voice pipeline for low-resource Nagaland creoles (Whisper, VITS, LLM

Article URL: https://www.ibtimes.co.uk/us-layoffs-skyrocket-highest-level-since-pandemic-tech-giants-blame-ai-40-cuts-1805380 Comments URL: https://news.ycombinator.com/item?id=48703722 Points: 6 # Comments: 1

Article URL: https://www.wsj.com/tech/ai/chinese-ai-anthropic-mythos-cybersecurity-574b02c2 Comments URL: https://news.ycombinator.com/item?id=48703592 Points: 6 # Comments: 3

Article URL: https://www.pentesty.co/blog/ai-super-bubble-cybersecurity-filter-2026 Comments URL: https://news.ycombinator.com/item?id=48703543 Points: 3 # Comments: 0

Article URL: https://www.spinnerrecruit.dev/ Comments URL: https://news.ycombinator.com/item?id=48703526 Points: 1 # Comments: 3

Article URL: https://jayacunzo.com/blog/your-move-chief Comments URL: https://news.ycombinator.com/item?id=48703452 Points: 35 # Comments: 6

<p>Driving a robot in simulation is easy. Bringing it into the real world? That's where the fun — and the headaches — begin.</p> <p>If you're building a custom 2WD differential drive robot using the <strong>Yahboom 4-Channel Encoder Motor Driver Board</strong> and <strong>520 DC Motors</strong>, you'll quickly hit a classic hardware challenge: how do you send velocity commands (<code>cmd_vel</code>) <em>and</em> read back encoder data (<code>odom</code>) over a single serial pipeline without cra

<blockquote> <p>Google open-sourced <code>DESIGN.md</code> — YAML tokens, a CLI linter, one-command Tailwind export. Great.But a format only helps people who already know what they want. What 0→1 products need isn't a format — it's a chain.</p> </blockquote> <p>Google Labs open-sourced something <a href="https://github.com/google-labs-code/design.md" rel="noopener noreferrer">interesting</a>: a standardized format for AI coding agents to read and write design tokens.</p> <div class="highlight js

<p>Every few months, someone says:</p> <blockquote> <p>“AI agents will replace developers.”</p> </blockquote> <p>I don’t think that is the right way to look at it.</p> <p>AI agents can write code, summarize logs, open pull requests, create tickets, call APIs, run tests, and even deploy software if we allow them to. That is impressive.</p> <p>But replacing humans is not just about doing tasks.</p> <p>In real production systems, the hard part is not only writing code. The hard part is understandin

Mythos 5 has been cleared for use by some trusted US organizations amid export controls.

Moumantai is a self-hosted runtime for creating personal "mini-apps": you define the app once, the server holds its state and logic (deterministic or powered by agent), and it can be accessed and rendered natively through a thin client on any device - phone, watch, browser, or an embedded board. Like many folks I was trying to build and customize little apps for myself, start by some web apps, but soon I realize I would want to access them across the devices I own, without opening the browser. W

<p>Sharing your health data with a cloud provider can feel like handing over the keys to your most private vault. Whether it's a persistent cough or a weird rash, the moment you hit "send" on a GPT-4 prompt, that data lives on a server somewhere. But what if your phone could think for itself? </p> <p>In this guide, we’re building a <strong>privacy-first health pre-diagnosis system</strong> using <strong>Local-first Health</strong> principles. By leveraging <strong>Edge AI</strong> and <strong>ML

<p>About a year ago I wrote about training a Pong paddle to move on its own. NEAT — NeuroEvolution of Augmenting Topologies. Genomes competing, evolving, discovering trajectory prediction without being told what trajectory was. I watched it work for a few minutes and moved on.</p> <p>I didn't know that was the first stop on anything.</p> <p>Last weekend I was four phases into a CartPole RL project when I looked back at the year and saw it. PyPongAI wasn't the beginning of a Pong project. It was

<p>There's a specific kind of confidence that a coding agent projects when it finishes a task. It doesn't hedge. It doesn't say "probably." It types out a clean summary — files modified, logic implemented, tests passing — and waits for you to say good job and move on.</p> <p>I burned weeks learning not to believe it.</p> <h2>The Session That Changed How I Work</h2> <p>It was a PrivyBot session — my personal autonomous AI assistant that runs on a home server I call Tower. I'd handed a phase direc

<p>AI is getting smarter fast.</p> <p>But there is still one very human problem most platforms have not solved:</p> <p><strong>What happens to the relationship, memory, tone, preferences, project context, and trust you build with an AI when you move to another platform?</strong></p> <p>Right now, most AI memory is trapped inside individual systems.</p> <p>You build context in ChatGPT.<br> Then you try Claude.<br> Then you use Grok.<br> Then you test Gemini.<br> Then a model changes, memory reset

<h1> Building SmartBasket: Find the Cheapest Groceries from a Receipt </h1> <p>Grocery prices change all the time, and comparing prices across supermarkets is a hassle. Most people won't spend time checking every item in multiple apps before shopping.</p> <p>That's why I built <strong>SmartBasket</strong>.</p> <p>SmartBasket lets users upload a photo of a shopping receipt, extracts every item using AI, compares prices across supermarkets, and recommends where to save money on the next shop.</p>