Wiki
Tags
About
FAQ
RSS
AIPOST

Fast, accurate AI news

An AI publication covering practical AI, AI security, performance, startups, health, ethics and industry news.

Latest articles

The most recent AI articles.

View all articles

Open-Weight Models Trail Frontier Labs by Months; Cheap Inference Is the Moat

Open-weight models typically match frontier performance within two to three months. Compute, not talent, is the main bottleneck for Chinese labs, says Nathan Lambert. Distillation buys a few months of progress but cannot replace base training. Lambert sees cheap, efficient inference, not secret algorithms, as the lasting moat. Teams with confidential data should weigh open-weight models on their own servers.

AI News

Doctor-Led AI Stack: One Tool per Job, From Guidelines to Consult Notes

One AI tool per task, with the doctor assigning work and checking every result. NotebookLM turns long clinical guidelines into summaries, flashcards and quizzes. Evidence search tools can fabricate references; read cited sources before acting. AI scribe drafts need review and correction right after each consultation. Start with a single tool for a week before adding the next.

AI Health

Claude Code and a Real iPhone: How a Teleprompter App Reached the App Store

Teleprompter app built with Claude Code passed App Store review after 215 prompts. Install every change on a connected iPhone instead of relying on the simulator. When the AI keeps guessing wrong about a bug, have it add diagnostic logging. Write automated UI tests early to catch regressions from later changes. Convert App Review demo videos to H.264 and compress them under the upload limit.

AI in Practice

New in each category

The newest article from each category.

Claude Code
Anthropic
AI agents

Claude Mods Turn Delete and Send Rules Into Code the Agent Must Obey

AI in Practice -

Mods intercept Claude's actions in code, so rules are enforced, not just requested. Requires Claude Code 2.1.287 or later in the terminal or desktop app, not web chat. Have Claude analyze your last 14 days of sessions to find repeated or risky work. Mods run unsandboxed with your full permissions, so third-party mods are risky. Build your own and check others with claude plugin validate before installing.

AGI
Anthropic
coding agents

AI Coding Agents Now Work Solo for a Day or Two, Shaping Anthropic's AGI View

AI News -

Anthropic researchers see AGI as realistically achievable within about two years. In 18 months, coding went from hand-typed lines to a day or two of solo agent work. Frontier models hit 40% to 60% on FrontierMath within about a year, up from 0%. Math and coding advanced first because answers there can be checked objectively. Hand agents whole multi-step tasks and have domain experts review their traces.

formal verification
AI agents
AI coding

Formal Verification for AI-Written Code: Why the Spec Comes Before the Proof

AI Security -

Writing a precise spec exposed a hidden Linux kernel bug before any proof began. Have the model draft a plan and review every line before it writes code. Use differential fuzzing to compare old and new versions when porting code. One firm's estimate: verification effort falls from double to 0.8 times with LLMs. Guard agents with a small verified monitor that enforces safe bounds.

GPT-6
OpenAI
Chess

OpenAI's New Astra Model Plays Real Chess, Then Falls to a Queen Sacrifice

AI Performance -

GPT-6 Astra played a full chess game with no illegal moves and reached a +3.5 edge. Missed key middlegame moves, then lost to a queen sacrifice and mate on h1. About 90 minutes for one game, with two to three minutes on hard moves. Its own post-game rating of its play was far below its real level. Test AI reasoning with positions just off known theory, not memorized lines.

AI startups
micro-SaaS
Claude Code

Non-technical founder's AI image tool hits $3,000 a month after cutting free use

AI Startups -

Non-technical founder's image-to-vector tool earns about $3,000 a month. Cutting free conversions from 25 to 3 started paid signups after zero revenue. About 1,000 AI-written pages in 12 languages lifted traffic tenfold in two weeks. Own scripts and self-hosted servers in place of a paid API keep margins above 80%. Launch early and let paying users' requests decide which features to build.

AI drug discovery
clinical trials
AlphaFold

AI in Drug Discovery: Smarter Clinical Trials Now, Not Instant Cures

AI Health -

Near-term AI payoff expected in trial patient selection and screening triage. Early cancer screens often yield 10 false positives for each true positive. Noubar Afeyan sees digital twins replacing human trials as still far off. Afeyan proposes regulating commercial AI agents rather than base language models. Judge claims of AI cures by clinical evidence, not promises.

Anthropic
Claude
model welfare

Could Claude Have Moral Status? Anthropic's Moves and the Case Against Them

AI Ethics -

Anthropic reportedly discussed Claude's consciousness with about 20 faith leaders. Updated usage policy bans sustained and needless abuse of Anthropic's models. Claude's constitution tells it to object to instructions it judges unethical. Core objection: consciousness cannot be proven or disproven by science. Businesses should check vendor constitutions and usage policies for refusal rules.

More to read

Recent articles worth reading next.

View all articles

Anthropic's Self-Replication Warning: What the 15 TV Box Test Shows

AI News -

Anthropic warns small AI models may soon copy themselves across home devices. Anthropic's test split a Google model across 15 cheap Android TV boxes. Technically possible, but slow and inefficient enough to look impractical today. The realistic threat resembles a botnet, where familiar defenses likely still work. Before trusting a warning, check the model, devices and security setup tested.

Read more

Copilot's Claude Pivot and Maia Chips Lead a Big Tech Reset of AI Strategy

AI News -

Microsoft added Anthropic's Claude models to Copilot after customer complaints. Microsoft plans several gigawatts of its own Maia AI chips starting next year. OpenAI is reportedly in talks to license Sora to a Hollywood-focused startup. Chan Zuckerberg Biohub unveiled a $1.8 billion AI partnership to fight disease. Test several AI models on your real work instead of relying on one provider.

Read more

Claude Lifts Zeta Zero Bound to Two-Thirds as Riemann Hypothesis Stays Unsolved

AI News -

Unreleased Claude proved over two-thirds of zeta zeros sit on the critical line. Best previous bound stayed just under 42% for 31 years after 40% in 1989. Not a proof of the Riemann hypothesis, and this method tops out near 68%. Checked in the Lean proof assistant, but journal peer review is still pending. In AI math news, check scope, verification, peer review and model disclosure.

Read more

Meta's New Agent Models Rank Second to Claude Opus 5.5 and Undercut GPT-6 Astra

AI News -

Muse Wonder Full ranks second overall at 55.8, behind Claude Opus 5.5 at 57.6. Test workload costs $0.80 on Muse Wonder and Claude Opus 5.5, $2 on GPT-6 Astra. Muse Code is a paid subscription with four plans from $5 to $200 a month. Move an existing workflow to the new model and compare answers before switching.

Read more

Browse categories

AI in Practice56 articles

How to put AI tools to work, on the job and in everyday life

AI News43 articles

Announcements, launches and policy changes across the AI industry, and what they mean

AI Security14 articles

Threats and vulnerabilities in AI systems, how defenses work, and how to use AI safely

AI Performance23 articles

Speed, quality and cost efficiency of AI models and tools, and how to read benchmark results

AI Startups16 articles

How to start and grow a business with AI, with revenue models and real cases

AI Health7 articles

How AI is used in medicine and health care, with its potential and limits

AI Ethics11 articles

The ethical and philosophical questions AI raises for people and society, seen through the humanities

View all articles

Topics

Get new articles by RSS or podcast

Add the RSS URL to your RSS reader or the podcast URL to your podcast app to get every new article as soon as it is published. The podcast reads out the headline and key points.

How to subscribe
AIPOST
© 2026 AIPOST. All rights reserved.

AIPOST is an AI publication covering practical AI, AI security, performance, startups, health, ethics and industry news. No account is needed, and our privacy policy explains how we handle personal information.