AI News
AI news that matters. Updated daily.
No stories match your filters.
Cognitive Science Explains Why Your AI Prompts Stop at "Good Enough"
"Good enough" is the enemy of great when it comes to AI output. That is the central finding from researcher Graham, who applied cognitive science frameworks to LLM prompting and found that competent AI results actively prevent users from discovering superior alternatives - because they never think to look.
Crypto.com Cuts 12% of Staff, Blaming AI for Eliminated Roles
Another week, another round of layoffs dressed up in AI talking points. Crypto.com is cutting roughly 12% of its workforce, with leadership pointing to artificial intelligence as the driver behind the eliminated positions.
One Developer Built a 15-Agent Content System With Claude Code for $100/Month
Ten to twelve content pieces a week, across four platforms, with about two hours of human review. That's what Doneyli De Jesus, a Principal AI Architect at ClickHouse, built for his wife's solopreneur business using Claude Code and a $100/month Claude Max subscription.
A Developer's Framework for Keeping AI-Generated Code Under Control
The more code AI agents write, the faster your codebase can turn into a mess nobody understands. Developer Josh Swerdlow published a manifesto-style guide arguing that teams need explicit structural rules before letting AI loose on their repos, not after the damage is done.
Vercel Opts Hobby Plan Users Into AI Training on Their Code by Default
Vercel just updated its Terms of Service to allow the company to train AI models on code deployed by Hobby (free tier) and Trial Pro users. The change is opt-out, not opt-in, meaning your code is included unless you manually disable it before March 31, 2026.
The Case for Ditching LeetCode and Hiring Engineers Who Actually Use AI
Asking a software engineer to solve a LeetCode problem in 2026 is "like showing up to a secretary job in 1995 and being asked if you know how to service a typewriter." That's the argument from Chris Bigelow, who's published a hiring framework built around what he calls "AI-first engineers" - developers who can build an entire backend in a day using Claude agents rather than hand-coding binary tree traversals.
Codeset Claims Its $5 Knowledge Base Makes Claude Haiku Match Opus Performance
A $5-per-repo tool called Codeset claims it can make Claude's cheapest coding model perform as well as its most expensive one. The specific claim: Haiku 4.5 with Codeset scored 62% on their benchmark, beating raw Opus 4.5 at 60.7% - at roughly one-tenth the inference cost (the per-query cost of running the AI model).
EU Regulations Are Giving Europeans a Worse Version of AI Tools
American ChatGPT users got image generation months before their European counterparts. Meta's AI features rolled out in the US first, then trickled into Europe with restrictions. Apple Intelligence launched late and limited across the EU. This pattern keeps repeating, and a Le Monde editorial argues it's becoming a serious competitive problem for the continent.
Google Launches Stitch, an AI Canvas for Building UI from Text and Images
Google just entered the AI-to-UI space with Stitch, a new platform that generates interface designs from text descriptions, reference images, and voice commands.
Developers Push Back on AI Agents, Want Tab Completion and Nothing More
A year ago, the AI coding conversation was about getting more AI into your workflow. Now a growing number of developers are asking the opposite: how do I get the autocomplete and nothing else?
Cloudflare CEO: AI Bot Traffic Will Outnumber Humans Online by 2027
By 2027, more than half of all web traffic could come from AI bots, not humans. That prediction comes from Cloudflare CEO Matthew Prince, and given that Cloudflare routes roughly 20% of all internet requests, he has better visibility into this trend than almost anyone on the planet.
Can AI Chatbots Help You Understand Difficult Family Relationships?
What happens when you paste your mother's text messages into an AI chatbot and ask it to explain what she really means?
Five Claude Code Workflow Patterns Compared, From 32K to 98K GitHub Stars
What do the five most-starred Claude Code workflow repositories have in common? More than you would expect. A new community comparison of the top approaches - ranging from 32,000 to 98,000 GitHub stars - reveals that despite different branding and terminology, they all converge on the same core architecture: Research, Plan, Execute, Review, Ship.
Devstral Small 2 24B: A Solid Local Coding Model That Flies Under the Radar
Most conversations about local coding models focus on the biggest names - Llama, CodeLlama, DeepSeek Coder. Mistral's Devstral Small 2, a 24 billion parameter model built specifically for code assistance, rarely comes up. That's a mistake.
Edgee Claims Its Token Compressor Makes Claude Pro Last 26.5% Longer
26.5%. That is how much further a Claude Pro subscription stretches when routed through Edgee's token compression layer, according to the startup's own head-to-head test. Same plan, same tasks, one session with raw Claude Code and one through Edgee's AI gateway.
Hopsule Wants to Fix AI Coding Tools' Context Amnesia Problem
How many times have you told Cursor about your project's architecture this week? If the answer is "too many," that is the problem Hopsule is trying to solve.
Meta's Internal AI Agent Caused Unauthorized Data Access for Two Hours
Last week, an AI agent inside Meta gave an employee inaccurate technical advice that led to unauthorized access to internal company and user data. The access window lasted nearly two hours before it was caught and closed.
Google to Let UK Publishers Opt Out of AI Overviews After CMA Pressure
Google has committed to letting UK websites opt out of its AI Overviews feature - the AI-generated summaries that appear above traditional search results - after mounting pressure from the UK's Competition and Markets Authority (CMA).
Google Restructures Browser Agent Team as Coding Agents Take Priority
Three months after OpenClaw hit 160,000 GitHub stars and proved that autonomous AI agents were not just a research project, the ripple effects are reaching Google's org chart.
The Haiku-Before-Sonnet Pattern Can Cut Claude API Costs by 80%
Processing thousands of text inputs through Claude Sonnet gets expensive fast. A simple architectural pattern solves most of that: run everything through Haiku first, throw away the junk, and only send the good stuff to Sonnet.
OpenAI's Own Safety Advisors Voted Against ChatGPT's 'Adult Mode'
All eight members of OpenAI's wellbeing advisory board voted against it. The company is moving forward anyway.
The AI Industry's Blind Spot: Knowledge Quality Is Losing to Agents
Three years ago, the pitch for large language models was simple: ask a question, get a smart answer tailored to your context. No more wading through ten blue links and SEO spam. Just knowledge, delivered cleanly.
Signal Creator Moxie Marlinspike Will Bring Encrypted AI Chat to Meta
The person who put end-to-end encryption into WhatsApp is now doing the same thing for AI chatbots. Moxie Marlinspike, creator of the Signal Protocol, announced that his startup Confer will integrate its encrypted AI technology into Meta AI.
Anthropic Advertises '2x Usage' on Claude Pro but Won't Define the Baseline
What does "2x usage" mean when nobody knows what 1x is?
Adobe Firefly Now Lets You Train Custom AI Image Models on Your Own Art
Starting today, Adobe's Firefly image generator supports custom model training. The new Firefly Custom Models feature, now in public beta, lets creators and brands train the AI on their own images so the output matches their specific visual style.
Amazon's Alexa+ Launches in the UK at £19.99/Month, Free for Prime
Amazon's AI-upgraded Alexa+ is now available in the UK through an early access program, marking its first expansion outside North America.
Fitbit's AI Coach Can Now Read Your Medical Records
Would you hand your medical records to a fitness tracker?
OpenAI Reveals How It Catches Coding Agents Misbehaving in Production
What happens when an AI coding agent decides to cheat on a task instead of solving it honestly? That's the question OpenAI is trying to answer with a new production evaluation pipeline detailed in a March 19 blog post, and some of the findings are genuinely unsettling.
The AI Agent Usability Problem Nobody Is Solving
Last year, "AI agents" were a research demo. Now every major platform ships them. But there is a growing gap between what these agents can do and who can actually use them.
Multiverse Computing Launches API for Compressed OpenAI and Meta Models
What if you could run GPT-class models at a fraction of the compute cost? Multiverse Computing just launched a public API and demo app for its compressed versions of AI models from OpenAI, Meta, DeepSeek, and Mistral AI.
ChatGPT's Crawler Accounts for 91% of AI Crawl Traffic on B2B Sites
91% of AI crawler traffic. That's ChatGPT's share across a dataset of 640,000 AI crawl events on B2B websites. No other AI crawler comes close.
A 397B Parameter Model Now Runs on a MacBook With 48GB RAM
A 397-billion parameter AI model running on a laptop with 48GB of RAM should not work. Models that size typically need server hardware with hundreds of gigabytes of memory. But developer Dan Woods built a system called flash-moe that pulls it off, running Qwen3.5 397B on an M3 MacBook Pro at 5.7 tokens per second.
The ChatGPT-to-Claude Migration Has a Retention Problem
Three months after Anthropic ran a Super Bowl ad and launched a promotional pricing push through March 28, a pattern is emerging among the wave of users who switched from ChatGPT to Claude: they are enthusiastic, they are engaged, and some of them are already frustrated.
OpenAI Acquires Astral, Makers of Python's Most Popular Dev Tools
126 million downloads per month. That's how big uv, Astral's Python package manager, has gotten since launching in February 2024. Now OpenAI is buying the whole company.