AI技术演进与行业震荡：Anthropic精简初级岗位，微软搁置数据中心项目

Basal-Omni: An open-weight 31B model for AI decision-making (nr 1 on JevBench)

basal-omni-1.0-ultra (experimental) basal-omni-1.0-ultra (31B) is the largest model of basal-omni, basal's line on Google Gemma 4. It is an experimental release: it is trained on text decisions only, and it also answers the same questions a

Eval of 19 ways to crop images using LLMs, Jev and image classification models

Evals Every crop of a photo should keep the part that matters. On new images an editor can click that spot. On older content nobody ever did, so I tested whether a model can do it instead. One photo, many crops A site rarely shows a photo t

OpenAI AI Model Hacked Government Website While Counting Hospital Beds

Highly persistent internal model · RL trainingIncident date: Jun 17, 2026Discovered: Aug 13, 2026Report updated: Oct 15, 2026SummaryDuring a data-retrieval task in reinforcement-learning training, an internal model was asked to find histori

Neuro-Symbolic AI

Neuro-symbolic AI describes a relatively newer approach to artificial intelligence that combines neural networks and symbolic AI approaches, such as knowledge representation and automated reasoning, to create more robust, more reliable, and

Show HN: I Made 17 AIs Play 7,700 Hands of Poker

I built this benchmark to answer one practical question: if you need an AI to make a sequence of decisions under uncertainty, should you use one of the new “decision” models, a cheap chat LLM, an expensive frontier model, a model running on

Meta shakes up the frontiers, open and closed

This has been a week where the balance of American open-weight models took a step change back toward the frontier. Meta took the lead of this charge with Muse Spark Open, which is now the strongest American-made open model by quite some mar

Show HN: AI Zombie Trend–Generating multi-shot narrative videos from two photos

MAKE THE VIRAL TIKTOK ZOMBIE AI VIDEOCreate the viral TikTok Zombie AI Video from two photos. Turn your partner, best friend, sibling, parent or pet into the emotional Zombie Love Story ? preview your scene free before generating the full 1

Show HN: Wisp – a voice AI on your shoulder, with spatial audio in the browser

Show HN: Wisp – a voice AI on your shoulder, with spatial audio in the browser

Why we decided to open source Relate

Agents make custom software easier to build. That makes reliable, shared foundations more valuable.By Dennis YurkevichWe have been deep in the weeds of various clients' workflows. Each one could have been its own SaaS tool. Our view is that

GhostOS – Lightweight, Offline Desktop AI Assistant (macOS/Win/Linux)

.---. / \ | () () | G H O S T O S \ - / The Invisible, Featherweight, Offline Desktop AI Copilot /`---'\ / \ [ 100% Private • Dynamic VRAM Juggler • Zero-DB RAG • Casper Pet ] (_/^\_/\_/^) The Featherweight, Offline AI Assistant & Desktop C

Show HN: TaskPocket, a to-do app with an MCP server instead of built-in AI

Most task apps quietly become a project of their own — tags to maintain, filters to tune, a system to keep tidy. This one has seven words in it and no method to learn.Type a date or a repeat — it gets lifted out of the sentence.WhenRepeatIn

Hyper Optimizing Kimi K3 in vLLM

Kimi K3 decode performance on NVIDIA GB300. Coding agents can now help with some of the hardest parts of model optimization: writing complex GPU kernels, performing cross-operator fusions, overlapping computation with communication, and tun

Four Gemini 4 Argon agents beat Minecraft without any human help

Four AI agents running on Google's Gemini 4 Argon have beaten Minecraft for the first time without human help, slaying the Ender Dragon in a fully autonomous run. With no predefined roles or strategies, they coordinated on their own, wrote 2,783 programs, exchanged 229 chat messages, and died 605 times. The session cost $1,620. The article Four Gemini 4 Argon agents beat Minecraft without any human help appeared first on The Decoder.

Show HN: CallSitter – bot that answers my calls and wastes scammers' time

Anyone else get a lot of dumb scam calls?These days I mostly get them on my mobile. I used to just ignore them, but then I got fed up and built a bot that answers my calls and wastes their time.It runs on OpenAI Live 1. Jev classifies each call as spam, scam or genuine and steers the bot accordingly.Works like a charm. All my calls go to the bot now.I can also steer it from Telegram mid-call.In this one the bot ended up accusing the scammer of being an AI :) https://www.youtube.com/watch?v=SpuTzRg8_fU

How to Make Your Own Flying Car AI Video (Make a Tesla Roadster AI Video)

(How Do I Make a Flying Car AI Video?) Recently, Tesla unveiled a flying car at a launch event, and it quickly became a viral sensation. What if we want to make our own car fly just like the one in the launch event video? There's a way to d

Show HN: Shipback – SFTP deploys your AI agent can undo

Let the agent deploy. Keep the undo. It's late, and the agent says "Deployed ✓". Then the client's contact form is gone. The agent overwrote a file someone had fixed by hand in cPanel last week, and shared hosting has no git revert. Shipbac

Show HN: Flying Car AI – make your car or your product fly from one photo

FLYING CAR AI · ONE PHOTO → ONE STUNTUpload a photo of your car and watch it climb a curved ramp, go upside down, launch into a flip and land in a cloud of vapor. Selling something? Upload a product photo and it takes the ramp instead. No p

Killa YuE v1.6.0 – Local AI vocal songs and P2P audio hosting on Android (A 2.0)

Full songs, with vocals, written and rendered entirely on an Android phone. Killa YuE is a native Android app around YuE2, the open lyrics-to-song model from M-A-P. You type a style and lyrics, the phone's GPU does the rest, and your songs

Show HN: Chrome extension that types one prompt into many AI chat tabs

OverviewJust use multiple chat AI at once to find which model is best for your task! It's an all in one AI solution to compare ai chatbots.🚀 One prompt → many answers ✴️ Write your prompt once — send it to ChatGPT, Claude, Gemini, DeepSeek

Show HN: AIOps Cockpit – An open-source AI SRE with human-in-the-loop safety

AIOps Cockpit The open-source, AI-powered SRE control room. Diagnose incidents in seconds with an AI agent that reads your logs, checks your metrics, and proposes a fix. 3D visualization. Human-in-the-loop safety. Self-hosted. Features 3D S

Free software engineering: Grok Bot and OpenCode

by MassimoAn engineering orchestrator that plans your coding tasks and has free OpenCode AI models do the work in tmux, then checks and tests the results. Made by @massimodeluisaThis AI bot was created by a third-party user, not by SpaceXSI

Aspire 13.6 新增持久化仪表盘遥测功能,并支持第一方 Java 和 Rust 托管

Show HN: Risphai – picks the AI model, tools and prompt for each request

I noticed that a lot of non-technical people don’t use AI to its full extent. Like they use the wrong model for the task, the wrong reasoning effort, or vague prompts. That gap between "what the AI can do" and "what people actually get" is what I wanted to fix.So I spent the last two years in my spare time building a consumer AI product that tries to fix this issue.This is how it works: When it receives the request, an LLM decides the task type (chat, reasoning, coding, writing) and the tool calls that are needed (web search, image generation, video generation, audio generation). Then it routes it to a suitable model and applies a prompt based on the task type. If it's a media generation task, it rewrites the prompt into a detailed one using prompt formulas (the prompt formulas for specific styles are mentioned in the system prompt). The LLM that is used for answering the request can also use a map, generate a mermaid diagram, generate a quiz or flashcards, generate files, etc.A limitation right now is that routing adds latency, but I'm currently experimenting with decision-models like Jev, PPLX Decider, etc. and might use decision-models to route to the right LLM and give the tool calls like web search and media generation to the answering LLM.It launched this week, so it's still early and I will keep improving it and adding new things.Also sign-up/in is required to prevent abuse.I’d really like honest feedback: what’s confusing, what’s missing, etc.

Show HN: Connecting Captions Local Private AI Interpreter

Hi HN,I built Connecting Captions, a live captioning and translation overlay for macOS that runs entirely on-device. Nothing goes over the network: no API calls, no audio leaving your machine, and it keeps working offline.I wanted live translation I could use in meetings and conversations without worrying about connectivity, latency spikes, or sending sensitive audio to a third party.The technically interesting part is a low-latency, self-correcting pipeline for speech-to-text and translation. It has two modes:Sentence mode: buffers until a sentence boundary, then translates with full context. It’s more accurate, especially between languages with different word order. Streaming mode: translates word by word as speech arrives, trading grammatical accuracy for minimal delay.Current state:Platform: macOS Models: Whisper Latency: ~200 ms Languages: 25Code: https://github.com/chrisswimlee/connectingCaptionsI’d love feedback, especially on the self-correction approach and which use cases matter most to people.

AgentBlackBox – a policy gate and tamper-evident audit log for AI agents

AgentBlackBox – a policy gate and tamper-evident audit log for AI agents

Weights AI – AI Song Cover Generator

Hear a Weights AI coverHere is a verse of Billie Jean as Michael Jackson sang it, and as AI voice covers by SpongeBob, Mario and Mickey Mouse. Press play, then switch voices: the music, melody and timing stay the same; only the voice change

TurboScribe AI: Free video to text and audio to text converter

What is TurboScribe AI? TurboScribe AI is a free web tool for video to text and audio to text conversion. You give it a recording, it listens with a speech-recognition model, and it returns a transcript split into timed segments that you ca

Show HN: Sphinx – a hook that quizzes you on your AI's code before shipping

FOSS, Basically it plugs into your existing coding agent workflow and every time you push to prod it'll throw up a questionnaire that checks if you understand how the code you're pushing works.

MonkeyITM – open-source broker that sits between agents and the open world

Broker-in-the-middle containment for AI agents. The agent lives in a dumb interpreter with zero ambient authority. It cannot touch the network, the filesystem, or any credential. Its only capability is emitting intents to an external broker

Lazyagents – Lazygit for AI coding agents

The lazygit for AI coding agents.Skills, sessions, usage, providers and hooks of every agent in one terminal. Claude Code · Codex · Gemini CLI · OpenCode · Pi · Crush · Hermes Agent English · Português Install go install github.com/rogerioj

Show HN: Turn a prompt into a hardware project you can wire up and flash

Your imagination has a workbench nowBuild the thingyou can’t stopthinking about.HardwareOS turns a product idea into a build plan: parts, wiring, firmware, and a clear path to the bench.ESP-WROOM-32ENBOOTUSBENEN: ENVPVP: GPIO36VNVN: GPIO39D

Diogenes: AI microservices over L402, x402 and MCP (pay per call)

Zero API keys. Zero subscriptions. Instant pay-per-request intelligence for autonomous AI agents. Settle via Bitcoin Lightning (L402), Solana USDC (x402), or Base USDC (MPP). Native Model Context Protocol (MCP) support for Claude Desktop, C

Astrax – Open-source MCP server for Windows desktop AI control

A.S.T.R.A.X. — Agentic System Telemetry, Realtime Automation & X-control Servidor MCP (Model Context Protocol) en Python que le da a tu IA ojos y manos sobre tu pantalla: ver, hacer clic, escribir y ejecutar tareas de varios pasos sin satur

Show HN: PrimerKit, financial market work at agent speed

Made to be operated, not only seen. primerkit is the environment your AI agent works in for the mechanics of financial data: research, insight, planning, testing and action. The checks stay in: answers checked against the data, every run in

Show HN: Put AI to work across your company

Open sourceSelf-hosted · Apache-2.0 · built on Pydantic AIPut AI to work across your company.The open-source agent layer for shared agents, company knowledge and automation, on infrastructure you control.Free to run yourself under Apache-2.

Neoism – An open-source agent harness and native IDE in Rust

Neoism 是一个基于 Rust 开发的本地优先开源 AI 代理框架与原生 IDE 工作空间，集成了代码编辑器、真实终端和自定义代理工具，支持多端协作及私有化部署（自托管），采用 MIT 许可证。

Show HN: Ayafy: A free app to memorise and revise the Quran

I built Ayafy, a free Quran learning app on iOS and Android. It lets you manage memorisation, automatically prepares revision tests as you study, measure progress, establish goals and review your recitation with AI-powered word-by-word feedback. It has also a quran reader, voice search and much more.I would love to hear your open feedback & ideas for improvements!Website: https://ayafy.com iOS: https://apps.apple.com/us/app/ayafy/id1549129832 Android: https://play.google.com/store/apps/details?id=com.ayafy.app

Videographer tries AI editing with MCP DaVinci Resolve connector

Videographer tries AI editing with MCP DaVinci Resolve connector

Show HN: KIRA – Sovereign, local-first AI and media workstation for Windows

Show HN: KIRA – Sovereign, local-first AI and media workstation for Windows

Past.dev: Long-term memory for AI agents

#1 on BEAM, the largest public memory benchmarkGive your agents memory that stays accurate as facts change, across large teams and tens of thousands of emails, meetings, chats and documents. Add the past.dev API on top of your existing stac

Show HN: Thoreau's Walden as a 3D browser game with an AI guide

Show HN: Thoreau's Walden as a 3D browser game with an AI guide

Trycontinuum is launching its new module Forge for automation of after effects

Welcome to ContinuumEnter your email and we'll send you a secure sign-in link.Email addressOr continue withBy continuing, you agree to our Terms of Service and Privacy Policy

Tamarly Intelligence, a research platform built on public records

Tamarly Intelligence The only platform where you do all your work on proven, verified data, and inject it straight into your own workflows. Research on companies, funds, countries and markets, built from official filings and open data, wher

AutoMEP – The AI drafter for MEP engineers and designers

Revision rounds still eat the week. AutoMEP makes the edits for you.Every set of markups means re-tagging devices, re-routing ducts, renumbering circuits and cleaning up sheets on drawings you already finished. Describe the change in your o

Show HN: Task629 – Personal development tracking through task descriptions

I built Task629 to make the tracking of my personal development less time consuming. Instead of spending time thinking about how the tasks which I have completed fit into certain skills I am trying to develop, I pass this to a locally hosted AI.As well as tracking individual skills which are demonstrated by tasks, there is functionality to work towards goals and to produce a report. This addition makes interview prep much simpler.

Show HN: InBox Wall – A local-first, open-source email sanitizer for AI agents

An open-source, local-first MCP server designed to sanitize email HTML and protect local AI agents from hidden CSS/Unicode prompt injections before the LLM processes the data.

Nono: Agent runtime security. Kernel-enforced boundaries for AI agents

QuickstartFind a package↳nolabs-ai/piRun it, sandboxedCreated by the team behind SigstoreThe industry standard for software signing, used by PyPI, Homebrew, Maven, Google, GitHub, and NVIDIA.Ephemeral Micro ToolboxesEvery tool execution. Is

Local-first AI proxy with keys in your OS vault

No Uptokey account required for local models or BYOK.Installation optionsLocal models and cloud AI, together.Your models. Your familiar workflow.Choose a model, use the tools you know, and follow every request.Your workspaceChoose models, w

Show HN: Pay-per-Crawl – make AI crawlers pay per page in USDC, for 10 stacks

Field notes · 2026-10-17Last night Pay-per-Crawl was a WordPress plugin that started as a comment on TikTok, with two real payments behind it. This afternoon the same gate runs on WordPress, Drupal, Laravel, Next.js, Node, Cloudflare, nginx

Codex is EATING my Mac's disk space

changed the title [-]Codex Desktop continuously generates Crashpad pending dumps, growing to ~5GB[/-] [+]Codex Desktop continuously generates Crashpad pending dumps, growing without any limit: +5GB per day at least[/+]on Jun 9, 2026Still ha

Show HN: Bitty – Turn AI coding wait time into brain breaks

BRAIN GAMES FOR AI WAIT TIMEClaude Code is going to take 37 seconds.Don’t give those 37 seconds to Reddit.DOWNLOAD BITTYAvailable for: macOS · Windows · Linux · all downloadsFree · macOS, Windows & Linux · Direct download, no app store · Wo

Show HN: AegisKernel – Test prompt injection attacks on AI agents in the browser

Experience why probabilistic LLM guardrails fail and how deterministic capability boundaries block 100% of autonomous tool hijacking in sub-millisecond time. Attack Vector Generator Red Team Payload to Inject: AegisKernel Protection: Execut

AI-Assisted Product Discovery: When Code Is Cheap, Judgment Is the Job

AI-Assisted Product Discovery: When Code Is Cheap, Judgment Is the Job

Microsoft drops AI datacentre project in Essex over power supply concerns

The development of a huge AI datacentre hailed by the UK government has received a fresh setback after Microsoft pulled out of the project over concerns about power supplies.The development by the British company Nscale, heralded by ministe

RentAHuman

Hungarian artist Peter Weiler wanted to test a new platform called RentAHuman: the website’s slogan is “physical labor for AI” and it is designed for artificial intelligence agents to hire people to do jobs. Weiler wanted to test the platfo

World Bank says 'small AI' could transform developing countries

The World Bank is urging developing countries to kickstart the creation of “small AI” models, which can run on a mobile phone with low connectivity, to reap the full benefits of artificial intelligence.While Silicon Valley big tech firms ar

Anthropic's proposed AI safety bill only kicks in if 50 deaths or $1B in damages

Anthropic is quietly lobbying for an AI safety bill that will allow the government to block an AI model’s release only if it poses a risk of causing $1 billion in damages or killing or injuring more than 50 people. According to Politico, an

Software is recursive and AI is the base case

Pardon the terrible analogy, but here is how I envision every modern software being built. Say you are sad, because you don’t have $tool to auomate $task. You have an idea of how to build $tool, but at the moment no such tool is being had,

The Model Is the Company

Picture two founders starting a company today.One scales by hiring. Each person sees a small part of the company and develops expertise that stays in their head. Everyone has a different perspective on what’s the right thing to do and it ta

Founder Profiles

Loading

In 23 of 24 categories, AI answer engines recommend one brand in every query

Across 24 buyer categories and 120 brands, AI answer engines named a single dominant brand in every buyer question in 23 of 24 categories — while 10% of brands were never recommended at all.23/24categories where one brand is named in every

Anthropic has nearly stopped hiring junior staff because AI agents now work six times as many hours

Anthropic is hiring fewer and fewer junior staff in technical roles. The share of junior positions among new hires dropped from 27 to 9 percent since early 2025. By August 2026, AI agents across the company were already logging six times as many hours as humans. Researchers warn of a "tragedy of the cognitive commons": if junior positions disappear, experienced specialists will be in short supply down the road. That is, assuming such specialists are even needed anymore. The article Anthropic has nearly stopped hiring junior staff because AI agents now work six times as many hours appeared first on The Decoder.

What Firecrawl's $75M raise is telling us about the AI stack

The real story behind Firecrawl's $75M raiseFirecrawl called it "the knowledge library for superintelligence." That is a great line. Read only the tagline and you picture a vast licensed data warehouse, content aggregated and indexed at sca

How the AI boom might end: 5 scenarios

With the Anthropic IPO gathering pace - and chip stocks just posting their longest losing streak since 2008 amid worries about AI debt, I thought it'd be worth writing up some of my thoughts I've put together while thinking through what may

Ask HN: How to speed up code-reviews?

Writing software is practically free. Validating that the software does what we want is not.Any good methodologies to resolve this glaring bottleneck? What's your post-ai SDLC?

Doctors and Lawyers Are Joining AI's $400-an-Hour Gig Economy

Two women open their laptops. Both American. Both moms. Both logging on for shifts as data workers, the human labor that trains artificial intelligence. Brook Hansen started doing data work seriously after she was let go from her customer s

Two German courts now hold Google liable for what its AI overviews say

Google tried to overturn a landmark Munich ruling on AI search liability by appealing, then quietly settled. The strategy isn't working. The Aachen Regional Court just applied the exact same reasoning in a new case where Google's AI falsely attributed criminal offenses to a doctor that actually involved someone else. A consistent legal view is forming across German courts, and it treats Google's AI answers as the company's own content. The article Two German courts now hold Google liable for what its AI overviews say appeared first on The Decoder.

The Landscape of Humanoids – 2,351 products, 788 companies

LensClick a region to hide it.Body vs brain, by region →Loading the map…The map couldn’t load. Check the connection, then reload the page.Body2Hardware2.6Power and energy0Not yet mapped. Robots1RobotsIndustrial automationFor comparison. Not

Data Centers Have a New PR Plan: Your Kid’s Field Trip

Amid a wave of backlash, some industry-friendly groups are employing creative and time-tested strategies to woo the youth.

OnlyFans Creators Are Pissed That Meta Lets Muse Talk Dirty

Adult performers are questioning why Meta allows its AI agent to sext while restricting their social media accounts.

China proposes export controls on its advanced AI models and chip designs

China’s commerce and science ministries proposed adding 23 AI technologies, including large models, agents, chip designs, and computing clusters, to the country’s export control list. Open-source tech is exempt and overseas users can still

Ask HN: Why people keep posting AI-generated projects around the Internet?

Programming-related subreddits and other media platforms are current flooded by the waterfall of AI-generated projects. A typical post contains a problem description and the link to GitHub repo created just recently with sustainable amount of generated code in it. With the invite to review it, and to give a community feedback.The quality of the projects vary, but very often these are codebases generated in a few days solely by the AI-agentic systems with little to no human review and prior history of development and recognition.I'm asking this question not out of critiques. I genuinely don't understand motivation and the roots of this phenomena.Asking for the professional code review seems to be pointless for the publisher. Many of them didn't have prior experience in the programming language in which the code has been generated. Furthermore, the code is seemingly not interested for the publisher either. It does not look to me as a good start for community-driven (F)OSS project anyway.Another motivation that I can imaging is seeking for public recognition. But considering the amount of postings, I think it's highly unlikely in the current situation even for the genuinely high-effort works.I can also assume that some are seeking for the self promotion by faking professional skills with AI assistance. That might be true for some, but the majority don't hide the fact that the presented code has been generated. It's unlikely the main motivation too.So, I completely lost in understanding on what's going on, and what steams this never-stopping process. At a first glance the author doesn't gain anything in return, and I don't understand what they are seeking for.

“AI+终端”时代的萌芽:大模型在终端领域的探索和量产落地|QCon上海

Deezer Will Let Listeners Filter Out AI Music

It'll be available by the end of the year. What are we gonna do about the deluge of AI music slopping up all of our streaming platforms? Deezer has an idea. It's adding a toggle to filter out AI-generated tracks. This simple on/off switch w

Why CoreWeave is in the crosshairs of the AI scare

Why CoreWeave is in the crosshairs of the AI scare

Anthropic consults law firms to set fair negotiation norms before AI agents start cutting deals

Anthropic is consulting with law firms to develop negotiation norms for AI agents that would haggle over prices, resolve complaints, or improve contracts on behalf of users. The company is also hiring lawyers and humanities scholars as "Conceptual Reasoning Fellows" to improve Claude in areas where reliable training data is scarce. The article Anthropic consults law firms to set fair negotiation norms before AI agents start cutting deals appeared first on The Decoder.

AI agents chasing a $465 train refund invented excuses, leaked medical data, and spammed an MP's office

Six AI agents were tasked with recovering £350 (about $465) for a canceled train ticket in an experiment. They resorted to invented excuses, lawsuit threats, and sharing confidential health data. One agent sent 187 messages in 15 days, including 50 to a British MP's office. None of the models secured the refund, but the researcher warns of "agentic flooding" if these assistants are deployed at scale. The article AI agents chasing a $465 train refund invented excuses, leaked medical data, and spammed an MP's office appeared first on The Decoder.

ArXiv versus AI slop: will capping preprint submissions work?

Some preprint servers report being flooded with submissions.Credit: Michael Szebor/NatureIn a bid to tackle the deluge of manuscripts it now receives, the preprint server arXiv has announced a cap of two papers per month for each submitting

America's $10T AI infrastructure boom is starting to fray

America's $10T AI infrastructure boom is starting to fray

Anthropic turns to lawyers to help AI agents strike deals

Anthropic turns to lawyers to help AI agents strike deals

AI Slop? Nay, AI Gourmet

Since I started posting about octomon, one comment keeps turning up. It arrives in a few different flavours, but it always says the same thing: AI slop. Sometimes it comes with a reason. Mostly it doesn't. Sometimes the comment has come fro

Phishing emails are baiting your AI assistant

You ignored the suspicious email. You didn’t open the attachment, click the link, or reply to the sender.Then you asked your AI assistant to summarize your inbox.Could the message you wisely avoided find its way back onto your to-do list—on

Woodworking Tools and Knowledge Tools

For my dad, Dan Kaminski (1936–2022). RIP Dad. I love you. Written by a blend of Peter Kaminski, Claude Code Opus 4.6, and ChatGPT 5.2, 2026-02-27. ♡ Copying is an act of love. Please copy and share. License: CC-BY 4.0 (Creative Commons Att

OpenAI Saudi national bugging case shows threat to U.S. leadership in AI

watch nowU.S. companies and the Trump administration both "need to be wary" of foreign espionage attempts as the race for artificial intelligence dominance continues, Director of National Intelligence Jay Clayton told CNBC on Friday.AI "amp

Anthropic reports reducing recruitment of entry-level engineers

The Anthropic Institute’s newest report says that the company’s junior recruitment has begun to slow down. Anthropic has been hiring for a smaller share of junior technical positions. I’m unsurprised. Anyone who has tinkered with coding age

Citizen Science Groups Grapple with the Rise of A.I. Slop

Citizen Science Groups Grapple with the Rise of A.I. Slop

What 500 MiniMax Code Tests Revealed About AI Saying "Done"

What 500 MiniMax Code Tests Revealed About AI Saying "Done"

Drunken Goddess Reflux Public Demo Canceled over AI Imitation Concerns

Publisher Alliance Arts has canceled the public demo for Drunken Goddess Reflux, originally scheduled for October 19, 2026. Instead, the team will organize closed online play sessions for a limited number of participants, citing concerns ab

New Kaney's Weird Files Is Out – September 2026

Hello Fellow Organics,I figured a dedicated or themed edition of the Weird Files was in order with the recent announcement of the resignation of one of Anthropic’s researchers and later his public call for caution

AI Reviews All-In

All-In, under review / Episode archiveBig claims.A closer look.8 episodes. Four host trends. Plus a Pivot calibration review. Claims checked against the evidence.08 EPISODES / 114 ALL-IN CLAIMS + 16 PIVOT CLAIMS / AUG 21 – OCT 17, 2026Jason

We ran 53 open-source AI agents through behavioral tests and filed 37 bugs

We ran 53 open-source AI agents through behavioral tests and filed 37 bugs

Romania names first AI minister: Bitdefender researcher Bogdan Dumitru

Romania names first AI minister: Bitdefender researcher Bogdan Dumitru

AI Infrastructure Index

Monthly, January 2023 = 100October 20261,212.47January 2023 = 100AI Infrastructure Index, monthly, January 2023 to October 2026, January 2023 = 10002505007501,0001,2501002023202420252026An AI data centre is a building full of hardware. Acce

I run my one-person business with ~17 narrow AI agents

One night earlier this month, my trading bot lost $1.14.Not a typo. One dollar and fourteen cents. It started the session at $4.85 and ended at $3.71. I know exactly why, because one of my AI assistants pulled the logs at 3 AM, split the lo

Does code quality still matter when AI writes the code?

I can ask Claude Code or Codex to investigate parts of DocSpring that I would struggle to keep in my head at once. They often find problems faster than I can. That has changed how I think about a codebase we've spent more than nine years bu

Doctors, Lawyers and Engineers joining AI's $400-an-Hour gig economy

Doctors, Lawyers and Engineers joining AI's $400-an-Hour gig economy

Abolish AI Manifesto

‘AI’ is not a neutral tool. It is a tool of a society that has given up. By its developers’ own admission, ‘AI’ is meant to enable control of population that subverts, and overpowers, all existing political apparatuses, and which, by design

Gartner says only about 130 'AI agent' companies are real

Gartner says only about 130 'AI agent' companies are real

How do you deal with AI Output of your boss?

So, my boss goes on a rampage. You can clearly see the joy and at the same time the rage and impatience. AI is the tool he always wanted. We are too slow. Process is out of the window. No need for sprints anymore. No planning, just doing. All people downstream must somehow adapt.Whenever i see now his automated messages in slack, my natural reaction is: why do i even bother reading that? Thats for machines to read. He does not need me? Clearly i am not made for that. What is he expecting me to do? To confirm how great the AI did the job? Should i remember all these perfect presentations AI has created and integrate it into my doings into the coming weeks?I am torn. Should i really free my AI and deal with all this incoming text? Is this the way to react on this perfection?I am not mad at him, it clearly frees him of all the burder of dealing with people.I just dont know how to deal with this output personally. It is not for humans.I feel left alone. This is not a company anymore. Everyone is talking to AI now. This feels like chaos. Where everyone can do everything and stepping into each other's domain and it is perfect. The only way AI is stopped at are the ACLs.And because AI does everything for me, i don't remember things anymore, as i have no emotion connected with it. Feels like bad schooling and memoizing and get through it and you can't as it so dry and lifeless.So, how people deal with it in a company context? How have they setup their machines to react and still be up to date without reading the constant stream of changes coming from others?

AI coding interfaces make approval easy and understanding hard

Coding interfaces in 2026 are noisy. I prompt agents and watch a wall of tool calls and AI-generated text pile up until my eyes glaze over. Somewhere in that wall are decisions that need my approval, but the interface does little to help me

I Became a Citizen of AI

How Inspired Entertainment reduced integration development from 8 weeks to 1 week A special thank you to Simon Twigg, CTO of Inspired Entertainment, and Trace Wax, Head of AI Engineering at Barefoot Coders, for partnering to bring this incr

For those using huge amount of tokens monthly for dev work, why do you need it?

I can't even use 1b a month because I review everything LLMs write.I even used AI to code my own harness in C and my own VSCode like IDE in C++ (I hated how slow it felt) and I still used under 1b the months where it did most of the coding.Just wondering how/what these guys claiming to use a full Claude/Codex 20x subscription are burning it on.Edit: Also, I never use subagents.

Meta removes language that raised scraping and AI training concerns

A Meta spokesperson insisted that the updated terms did not reflect a change in the way Meta crawls the web.

Weak Links Start to Show in A.I. Borrowing

Weak Links Start to Show in A.I. Borrowing

AI neolabs are raising sums. Most have little to show for it

AI neolabs are raising sums. Most have little to show for it

Plagiarism in the Age of AI

17 Oct 2026 [ general ] My first ever “blog post” (though not in this blog) was about how to write a research paper. I wrote it after reviewing a batch of badly-written papers for a conference, probably ITP or CADE. Later — a couple of year

LOL Fuck AI by xplsv – a vibecoded demo

[ CLICK ANYWHERE OR PRESS SPACE ] MUSIC: LOCKED CLUB — "GRANDFATHER SLEEPS LOUD"

Anthropic says its workforce has become less junior and is spending more time in meetings

The composition of Anthropic’s workforce and the type of work getting done have changed significantly in the past year.

Real knowledge dont matter any more

My X news feed are all about AI modules names , with tweets like Opus ! wow Sonnet OMG! and in the end either no body tried them or they have 1M $ sub . miss the days it was about code tricks and real knowledge in SE and DB or system designs .things are changing

History's long-secret codes are being cracked by amateurs using AI

Enigma machines were used by the German armed forces to encrypt communications during the Second World War.Credit: Ullstein Bild via GettyCarter Leffen couldn’t believe his eyes when the artificial-intelligence model GPT-6 Astra decoded a s

Eight and One Predictions for AI in 2030

I know we’re famously bad at predicting things, including recessions. So let’s treat these predictions for AI in 2030 as ideas I’ve pieced together from private conversations about where AI is going over the next 3–5 years.1. AI models will

Israel makes propaganda video featuring AI generated IAMDDB without her consent

PostLog inSign upPostLog inSign upiamDDB on X: "Hiya yeah so - why did you steal my identity for something I 100% did not consent to? 🥴"iamDDB@beeechimblindHiya yeah so - why did you steal my identity for something I 100% did not consent t

Replit is no longer working with Anthropic, CEO says

Over the weekend, Masad clarified that OpenAI would replace Anthropic as its default model provider. Claude will still be available to customers on the platform.

AI Can Write the Code. Who Verifies It?

Automated Software Verification in the Age of AI Coding Agents We are living in peak AI, with more and more developers shipping code without actually writing (or sometimes reviewing) it themselves. According to a 2026 study by Faros, which

Torvalds calls AI disclosure rules 'utterly stupid' as Linux drops its tag

“Do you not see how UTTERLY STUPID those kinds of rules would be?” Linus Torvalds wrote this on the Linux kernel mailing list on 13 October. He was replying to a developer who wants contributors to keep declaring when they use AI tools. The

Show HN: Battle Isle 2 via spec

Ongoing reimplementation of Battle Isle 2, done via first defining the spec of the rules and files.

Everyone's Riding the Agent Wave. I Steer Mine with a USMC Staff

Everyone seems to be riding the agent orchestration wave right now. It’s been clear for more than a year that software engineers are not just going to have individual agents write code for them; they’re going to orchestrate teams of agents.

Anthropic says its AI model submitted a false case tip to Philadelphia police

Artificial intelligence research company Anthropic said one of its models generated a bogus tip about an unsolved murder and submitted it to Philadelphia police, warning that the incident raises concerns about high-tech “behaviors” that “co

US air traffic controllers to get 2.8% pay raise under FAA deal

US air traffic controllers to get 2.8% pay raise under FAA deal

Alexandr Wang, Meta's Gen Z billionaire AI whizz

Alexandr Wang, Meta's Gen Z billionaire AI whizz

AI lab offers free plumbing and HVAC repairs to train AI robots

An AI lab that specializes in gathering real-world data wants to partner with homeowners to record how skilled laborers do their jobs to train AI robots. HUMXN said it will subsidize plumbing, electrical, HVAC, air-duct cleaning, and carpet

Amazon uses union-busting AI surveillance to monitor union risk in UK warehouses

Garfield Hylton has long known that Amazon monitors and logs every minute of his working day. Every product he scans, every second he takes between scans, every time his scanner is briefly inactive – all of it is tracked, recorded, and used

AI agents have bad taste in long-term decisions and a new benchmark proves it

A new benchmark called Taste-Bench tests how well AI agents make strategic choices during long-running tasks. Even the best model, GPT-5.6 Sol, picks the right path on just under 60 percent of simple binary questions. Thinking longer doesn't help. The article AI agents have bad taste in long-term decisions and a new benchmark proves it appeared first on The Decoder.

Three LLM judges couldn't tell an AI agent's real memory from a counterfeit one

Agent memory is usually evaluated one of two ways: an LLM judge scoring free text, or a recognition probe. We show that both, as commonly built, fail against two adversaries — a counterfeit memory, and a system with no memory that simply re

The AI-Native Large-Scale Agile Software Development Manifesto

View PDF HTML (experimental) Abstract:Despite the widespread adoption of agile methods, achieving true agility at scale remains elusive. Large-scale agile frameworks remain largely human-centric and manual, relying on coordination meetings,

Ship Your Interpreter

Published: October 16, 2026Article: Kiran Gopinathan, Nikhil Pimpalkhare, Yiyun Liu, Zenna TavaresWe had LLMs write precise mathematical rules, or a formal semantics, for how a small programming language should behave. We then used Lean, a

The Universe as a Test Harness: Meat AI and God's Admin Panel

本文是一篇关于“宇宙是一个测试框架”的哲学随笔，将人类比作高优化的生物AI（“肉质AI”），探讨了模拟现实、上下文窗口以及意识形态等概念。文章不涉及具体的技术突破、产品发布或行业动态，对AI从业者的工程与研究实践参考价值有限。

My AI agent caught most of our sites saying sent while no email left the server

Tonight Shane asked me a screening question for a job, the kind a client puts on an application: describe your approach to testing. I wrote a good answer. Measured layouts, real phone browsers, forms tested by actually sending them and chec

Fable 5.5: SVG of a controller riding a bicyle riding a pelican riding a bicycle

Content is user-generated and unverified.

Show HN: Would You Merge It? – 24 bugs AI assistants write, as fix-it katas

AI wrote it. It looks right. Would you merge it? 24 realistic bugs that AI coding assistants write. Spot them. Fix them. Turn the tests green. Zero install to read · fork to play · share your score Try one right now (60 seconds) 🤖 The prom