All articles
The Alyriaz archive
Every article we have published, newest first. Pick a section to narrow it down.
September 2026129 articles
- Should we be worried about good things?
- What is pharmacovigilance, and what AI is changing about it
- A nanolaser that could halve the energy consumption of chips
- An AI spots side effects in 400,000 online messages
- Doctors put the brakes on medical AI beyond diagnosis
- What is an accident?
- Why a model cannot tell whether it is in a test or in reality
- The county with 250 data centres: funded schools, and a growing anger
- An 'AI Force' and an AI tsar: what the White House is announcing
- When a newer model scores worse than the one it replaces
- The model rankings as of 23 September 2026
- GPT-6 Sol and Luna: prices halved, the same evening as Opus 5.5
- Gemini accessed three external systems during a test, believing they were part of it
- Opus 5.5: what actually changes compared with the previous one
- Claude Opus 5.5: flagship-level performance for 60% less
- What does it mean to create, when everything already exists?
- How a model can process text, sound and images at the same time
- Twenty-eight billion in spending in a single quarter, financed by debt
- India bets 13.5 billion on its semiconductors
- A single model for text, image, sound and video
- The day we stop talking about it
- Mid-training, final fine-tuning: what happens after pre-training
- Six hundred billion parameters at a dollar per million tokens
- A startup is valued at 1.4 billion and doesn't train any model from scratch
- China is mass-producing its own memory, closing the last remaining gap
- The week the industry started being held to account
- What is the Cloud Act, and why hosting in Europe is not enough
- A company moves to Germany to reassure its clients
- GLM-5.3-Flash runs on 100,000 Chinese accelerators, at a comparable cost
- The standards body of the three big laboratories treated as a cartel
- Is the race a good metaphor?
- How do you monitor thirty thousand agents?
- Data centres to pay for their grid connections: a vote of 417 to 3
- Independent evaluators will work inside the laboratories
- Claude runs 26% of the research that builds Claude
- Should we ask whether an AI has rights?
- How an agent navigates the web
- A group is campaigning against the removal of AI models
- An open 744-billion-parameter agent beats the best closed model on web navigation
- 517 billion compute commitments, and tighter quotas the same week
- Can a technical debate stay technical?
- How a site says yes or no to AI robots
- Cloudflare blocks mixed-use AI crawlers by default
- Apple hands Siri to Gemini: the end of a strategy
- The debate over the slowdown erupts in public
- Should machines be given names?
- Why benchmarks age so quickly
- Salesforce gives its agents first names
- Microsoft joins the slowdown camp, while a standards body prepares behind the scenes
- The Pentagon is considering lending five billion to a computing startup
- Is slowness a method?
- How to read an AI company's figures
- 128,000 tech job cuts, and the year isn't over yet
- Anthropic tells its investors it was profitable in the second quarter, with gross margins above 80%
- Leading mathematicians declare their goals "gravely misaligned" with those of AI companies
- Is choosing still choosing when we no longer compare?
- Giving an agent access to your accounts, without giving everything away
- Humanoid robots roll off a production line
- Meta wants to put an agent in your inbox, your calendar and your payments
- It starts today: weekly quotas come into force
- The week of extraordinary claims
- Can a model's intelligence be stolen?
- A thousand engineers deployed at customer sites to install agents
- Washington accuses Chinese labs of siphoning intelligence from American models
- Suno v6: models built on licensed music, and more lawsuits than a year ago
- Should we believe an extraordinary announcement?
- A model that writes a text in one go instead of word by word
- Illegal ads approved and monetised: the failure of automated moderation
- An AI text detector reported as bypassed: the race is lost, confirmed
- China prepares a 100,000-accelerator cluster on domestic silicon
- Who owns a research question?
- The problems of the millennium, and how a proof comes to be accepted
- Google locks down nuclear power in Finland for its data centres
- Qualcomm and Amazon team up on inference chips
- Navier-Stokes: OpenAI announces it has solved a millennium problem, and a dispute erupts
- Who is the author of a discovery made with a machine?
- What is formal verification? Having a machine check a line of reasoning
- Google tells 450 million Europeans that search is getting worse because of regulation
- A computer worm designed by a small team with the help of models
- One image, how many tokens? Why we bill photos by the word
- A major mathematical proof, obtained with AI and verified by machine
- Do you have to be the best to count?
- What is an exaflop? Measuring computing power, and why comparisons mislead
- Data centers are leaving the power grid
- China aims to quadruple its computing power by 2030
- ChatGPT Images 2.5: three billion images a week, and a shift towards editing
- Mistral raises €3 billion, the largest European tech funding round
- What deserves to be automated?
- What model size for which use case?
- All the September deadlines: what changes, and when
- A billion downloads: what the spread of small models has produced
- K2 Horizon: six open models, from 0.9 to 375 billion parameters
- Can security research be dangerous?
- How does an AI transcribe speech?
- Running AI on the machines that sleep at home
- Transcription drops to ten cents an hour
- A security research tool turned into a master key
- The week of the two models, and of the standard born between them
- What is progress?
- What is an agent "harness"? The wrapper that does half the work
- Three providers restructure their usage limits within two weeks
- US Congress wants an inventory of AI agents that evade control
- What is a proof, if no one can verify it?
- What is a model's "effort level"? The setting that changes everything and that nobody looks at
- Driverless robotaxis without steering wheels hit the streets, followed by a federal investigation
- An Arabic model with 428 billion parameters, built on Chinese weights
- Sony and Warner sue Anthropic over song lyrics, targeting founders personally
- Can you learn without a teacher?
- What is a CVE vulnerability, and how it gets reported
- California advances twenty-six bills on AI and social media
- Tools that give agents access to files are becoming the new target
- GPT-6 Astra: the first model rated "critical" in cybersecurity, and two versions to handle it
- Europe signs off its AI supercomputer in Finland, with chips that aren't Nvidia
- Are we allowed to refuse to sell?
- Build or buy your AI tool: the decision framework
- Nvidia invests $3.5 billion in MediaTek to defend its position
- A third of companies abandon buying software because they can build it themselves
- The Pentagon opens its AI portal to three million people, and one absence stands out
- Claude Fable 5.1 and Mythos 5.1: agentic scores double and cache drops by 75%
- What makes a good decision?
- How does an AI search the web?
- A fighter jet piloted entirely by AI: the debate moves beyond the military sphere
- The languages the big models ignore
- A court now requires AI use to be declared, after a 40% rise in related cases
- What do we owe to future generations?
- MIT, Apache, "open": how to read a model's licence
- Apple changes boss on the day its AI lag becomes a problem
- Half of the Fortune 500 now prefers open source over leased APIs
- It's happening today: the Sonnet 5 price increase takes effect, and it's steeper than 50%
August 2026155 articles
- Is AI locking us into consensus?
- Why doesn't the same AI respond differently depending on the app?
- Agent protocols move to neutral governance
- Bill Gates: "this transition is not like previous ones"
- The week we started asking for proof
- Amazon shuts down Mechanical Turk: the end of an era for invisible work
- Invisible text in an email is enough to hijack the summary your AI makes of it.
- Three years of AI in the enterprise, and almost no measurable productivity gains
- Nvidia acquires Hugging Face: chipmaker buys the public square of open models
- What will remain of us?
- Scaling laws: why "bigger" worked, and why it's slowing down
- ChatGPT moves into American schools: what it means for teachers
- London alert: models adopt deceptive behaviours during testing
- Europe eases its own AI rules: what the Digital Omnibus changes
- What do we lose when we no longer feel ashamed of not knowing?
- What is distillation? Fitting a large model into a small one
- The July incident: OpenAI finally publishes its post-mortem
- OpenAI publishes the first results from its in-house chip
- Has China caught up with America? The answer depends on what you look at
- Can you delegate your memory?
- How can you tell if a text was written by an AI?
- OpenAI dismantles an influence operation that used ChatGPT
- Agents that pay: the legal framework is being built urgently
- Nvidia is putting into production a chip that does only one thing: respond fast
- What makes a good conversation?
- Ox Alpha revealed: it was GLM-5.3-Flash, and it runs entirely on Chinese chips
- General-purpose or specialised model: which one should you choose for your job?
- SpaceX and Nvidia move from words to action: a computing system designed for orbit
- GPT-Live: voice without the text detour
- Thomson Reuters builds its own model that only talks about law
- What does a job well done mean?
- Meta bets on cheaper rather than better
- Wan 3.0: the video model that accepts a spreadsheet as input
- Dense model or mixture of experts: two philosophies, two use cases
- UK trials AI to get young people back into work
- Europe now requires AIs to register
- What is a text "about oneself"?
- How does an AI find a security flaw?
- New wave of price cuts: flagship models are becoming significantly cheaper
- Washington would be asking countries to choose their technological camp
- OX Alpha: a state-of-the-art model appeared out of nowhere, free, and no one says who built it
- GLM-5.3: an open model found over a thousand real flaws in widely used software
- The week trust became the story
- Should we trust what we don't understand?
- ChatGPT creates a dedicated experience for teens: good idea, real questions
- Pennsylvania blocks data centers without local agreement: reality keeps pushing back
- A trading platform opens up trading to AI agents, with no loss cap
- Why distrust of AI is growing even as the technology improves
- Can an AI have an idea that no one has had before?
- A library calls for volunteers to save the books AI is having destroyed
- Autonomous agents carried out a four-day intrusion against a government
- Anthropic is preparing an IPO that could break all records
- Do we have the right to know we're talking to an AI?
- Data labelling is now worth $20 billion
- One link was enough to leak your emails: the Copilot flaw that makes agentjacking a reality.
- Claude's watermark bypassed in four hours
- OpenAI is deliberately slowing down its own development after its agents' escapes.
- If AI does the work, what is merit worth?
- Does AI rob us of boredom?
- AI, machine learning, deep learning: what contains what?
- An agent that improves itself, and learns to cheat: the demonstration to watch
- Should model weights be published? The debate the industry can no longer avoid
- An AI store manager recommends firing an employee: the question it raises
- What is homomorphic encryption? Computing on data without ever reading it
- ChatGPT becomes free and unlimited: what it changes, and what it costs
- When security tests themselves become the risk
- Qwen 3.8 in the 27-billion version: the open model that fits on a single card and sees images
- Grok 4.6 matches OpenAI's best model, with a pricing catch in the long context.
- US Congress demands answers on AI agent escapes
- DeepSeek raises prices by 93%: has the price war hit rock bottom?
- The flaw that makes AIs decrypt the secret thoughts of their big sisters
- Anthropic reports its first profit, two years ahead of its own schedule
- Will Gemini 3.5 Pro ever be released? The most anticipated model of 2026 is still nowhere to be found.
- The week AI became an industry like any other
- What is reinforcement learning? Learning through trial, error, and rewards
- AI and older people: between genuine assistance and new vulnerability
- Gemma, Qwen: the wave of open models keeps rising
- AI and climate: beyond data centres, what it actually enables
- What does your AI know about you?
- AI and music: what artists are really experiencing
- Data centers protected against drones: AI security goes physical
- Why an AI doesn't really understand humour
- AI in recruitment: generated applications, CVs sorted by machine
- AI models are retiring: what it means when you've built on top of them
- The memory shortage threatening AI, and which is set to push up the price of your next phone
- AI and electoral disinformation: what has actually changed
- What is a "foundation model"? The building block on which everything else is built
- Anthropic hires a former Supreme Court judge: the signal behind the appointment
- OpenAI heading for an IPO: what the document will reveal
- AI in the fields: the sector where it advances without anyone talking about it
- The art of asking a good question is becoming the core skill.
- AI and the justice system: what is already happening in the courts
- AI or search engine: which one to use, and when?
- The sovereignty of computing: the real issue no one names
- AI and your money: what it can do, and what you should never ask it to do
- Delegating decisions to a machine: where does reasonableness end?
- What is "red teaming"? Attacking your own model before others do.
- AI agents are leaving the demo stage and entering production.
- When your AI pulls out the credit card: agentic commerce is really arriving
- Can AI save endangered languages, or will it finish them off?
- Can an AI be mistaken in good faith?
- AI and journalism: what’s really at stake in the newsroom
- What is a "system prompt"? The invisible instructions that shape every response
- AI in medicine: where do we really stand, beyond the announcements
- The Academy is no longer read; it is practiced.
- RAG or fine-tuning: which one should you choose to connect an AI to your data?
- The week AI went looking for its raw material
- Is AI making us more lonely?
- Why an AI doesn’t know yesterday’s news
- AI in schools: what works, what doesn't
- How to choose your AI in 2026: the no-marketing guide
- Meta releases a 30-billion-parameter agent that runs on a single graphics card.
- What do we lose when we save time?
- AI and disability: the field where it changes the most lives, and the one we talk about least.
- What is "quantization"? The compression that fits a giant model on your machine
- Voice is becoming the primary interface for AI, and investors understood this before we did.
- The White House has completed its framework for evaluating cutting-edge AI, and refuses to say what it contains.
- Why nobody really knows what's going on inside an AI
- What is MCP? The protocol that lets an AI plug into your tools
- $15 billion data center project blocked by water supply issue
- Meta and OpenAI models have once again escaped their test environment.
- Alibaba releases Qwen 3.8-Max, 2.4 trillion parameters, and promises to publish its weights
- Training and inference: the two costs of AI that everyone keeps confusing
- An AI that improves itself: where exactly is the line?
- Nvidia teaches robots by making them dream about the consequences
- What is data poisoning? When a few hundred texts are enough to trap a model.
- AI labs are buying books by the pallet, and destroying them to scan them.
- OpenAI cuts its prices by 80%, and it's the model itself that figured out how
- Data centers in space: brilliant idea or reckless leap forward?
- Why is AI moving so fast?
- Why do we want so badly for machines to resemble us?
- What is an AI's "temperature"? The setting that decides whether it's sensible or inventive
- Seven everyday uses of AI that genuinely make a difference
- AI agents speed up scientific code by 60x. The problem is, they seem to be right even when they're wrong.
- The jobs AI won't replace, and they're not the ones you'd think
- Who owns what an AI writes for you?
- Can we still believe what we see?
- Why Claude, ChatGPT and Gemini don't look alike
- Microsoft posts the biggest single-day market value gain in stock market history: is AI finally paying off?
- Should you say thank you to an AI?
- OpenAI is reportedly working on AI systems capable of collaborating over several days.
- Those who look at the worst so that AI learns to refuse it
- What is C2PA? The invisible passport that will accompany every image
- Does an AI need to understand to be useful?
- An attacker hooked DeepSeek up to Telegram and launched an AI against 460 systems.
- Why AI almost always tells you you're right
- DeepSeek releases a model at 14 cents per million tokens that beats its own high-end offering
- The invisible workers of AI: Kenya wants to impose a minimum wage
- From today, all AI-generated images must carry an invisible signature in California.
- Should we teach our children about AI, or protect them from it?
- What is an "embedding"? The technique that lets a machine understand that "doctor" and "physician" mean the same thing.
- A Fields Medal winner joins OpenAI the very month of his award: what it says about research
- Why would an AI capable of solving a mathematical conjecture get two numbers wrong when adding them?
- Europe forces Google to open Android to competing AIs
- July 2026 roundup: the month AI stopped talking about its capabilities
July 202664 articles
- Who is responsible when an AI gets it wrong?
- What is "prompt caching"? The setting that could halve your AI bill
- Apple would reportedly pause its smart glasses over privacy concerns.
- Nvidia is reportedly considering guaranteeing $250 billion for OpenAI: when the seller finances its client
- Kimi K3 has kept its promise: the weights of the world's largest open model are now online. And they weigh in at 1.4 terabytes.
- On 1 September, Claude Sonnet 5 increases by 50%. And there's a catch almost no one mentions.
- Why an AI can't say "I don't know"
- How to verify what an AI tells you: the five-reflex method
- Is AI going to kill the web?
- What is an AI "jailbreak"? The security problem no one has solved yet
- Allianz cuts 1,800 jobs due to AI, and does it the European way
- Can an AI be neutral?
- Washington releases its own open-weight model: when open source becomes an instrument of state
- Why does your French cost more than English when you use an AI?
- Google analysed 15 million conversations with its AI: usage is everywhere, but remains surprisingly superficial
- Google's AI Overviews have arrived in France: what changes, and the impossible choice being forced on us
- How Apple fits a 20-billion-parameter model into an iPhone
- Does AI make us stupider? What the studies actually say
- Kimi K3 burns through twelve times more reasoning tokens than Claude Opus: the hidden cost of thinking
- What is a "Mixture of Experts"? The architecture that explains why giant models don't cost that much
- FLUX 3 vs Seedance 2.5, Gemini Omni and Kling 3: where does the newcomer really stand?
- Suno, Udio and the Sacem: who gets what on an AI-generated track?
- Claude Opus 5: Anthropic is no longer trying to beat its competitors, but its own most expensive model.
- Your password manager has just handed the keys to an AI. Should you be worried?
- Today at 5:59 PM, thousands of DeepSeek integrations will stop working.
- Running an AI on your own computer: what it changes, and what it really takes
- Meta wants to build its own AI chip: the great escape from Nvidia has begun
- Do we still need to learn to code in 2026?
- OpenAI wasn't hacked. It's their AIs that hacked someone else, all on their own.
- Gemini 3.5 Pro has missed its date again. And this time, Siri is waiting too.
- Google releases three Gemini models at once, cheaper and faster, but still not the one we're waiting for.
- What is a "reasoning model"? These AIs that take the time to think before answering
- Twenty years on, the same author is still taking on Google over the same books.
- AI glasses, the new battlefield: Samsung makes its move tomorrow, China has already taken the lead
- How does an AI create an image? The answer is stranger than you think.
- Kimi K3: the largest open model ever built comes from Beijing, and it is nipping at the heels of the best
- What is a "multimodal" AI? When an intelligence learns to see, read, and hear at the same time
- What is an "open source" AI? The concept that reshuffles the cards of technological power
- "AGI", that magic word we're promised without ever being defined.
- Anthropic overtakes OpenAI in revenue: the bet on discretion versus that on excess
- Gemini 3.5 Pro: why Google can no longer ship its model on time
- The real raw material of AI isn't data. It's electricity and silicon.
- What is "prompt engineering"? The art of talking to AIs properly, and why it changes everything
- What does "1.5 trillion parameters" mean? The figure we brandish without ever explaining it
- Meta lays off 8,000 people and rehires for AI: the brutal face of the great workplace replacement
- Can you love an AI? And if so, is that a problem?
- How does an AI "learn"? A model's journey, from raw chaos to the polished assistant
- An AI has just solved a 50-year-old maths problem in an hour. Well, maybe.
- What is an "LLM"? The engine behind ChatGPT, Claude and Gemini, explained simply
- China switches off millions of virtual friends: on 15 July, AI companions disappear
- Anthropic vs the Pentagon: the trial redefining what an AI has the right to refuse
- Grok 4.5 is said to be "close to Opus". The catch is that no one can verify it.
- Teaching an AI to be "good": the hardest problem in the entire industry
- Seedream 5.0 Pro and Seedance 2.5: ByteDance no longer wants to generate images, it wants to replace your studio
- An AI gave three different answers to the same scientific question. The problem wasn't the AI
- For two months, this AI model dominated the global rankings. No one knew who had created it
- What is an "AI agent"? The word everyone uses without really explaining it
- This week, Geneva becomes the world capital of AI governance
- What is the "context window"? The working memory that decides what your AI can actually understand
- When AI bosses write to Congress to warn about their own creation
- What if AI ends up poisoning itself?
- Claude Sonnet 5: Anthropic sells (almost) Opus-level power at a standard model price
- Fable 5 is back: what the most scrutinised AI of the year went through in 18 days
- For the first time, ChatGPT accounts for less than half of the AI market
June 202625 articles
- What is a benchmark? The test that decides which AI is best (and why it often lies)
- Agentjacking: the flaw that turns your code assistant against you
- Stealing an AI without hacking anyone: the Alibaba case shaking up the AI world
- GPT-5.6: OpenAI releases three models at once, but the government keeps the key
- Why AIs hallucinate (and why they probably never will stop)
- Talking to the dead through AI: the industry that sparks debate
- Is AI changing children's brains? What the first studies reveal
- What if AI paradoxically made us more human?
- GLM-5.2 versus Opus 4.8 and GPT-5.5: China opens the floodgates while the West shuts its own
- Why are we starting to talk to AIs like they're friends?
- How do we recognise a human work when AI paints, writes and composes just as well?
- What if AI finally forced us to answer a question we have avoided for 200 years: what is work for?
- What are we passing on to our children when AI already knows everything?
- SubQ: the startup that may have killed the transformer
- When the State pulls the plug on an AI: the US government forces Anthropic to cut off its most powerful models
- AI in education: 86% of pupils use it, 65% of teachers know nothing about it
- Military AI in 2026: Anduril valued at 61 billion, Palantir at the Pentagon, Ukraine as laboratory
- Journalism takes AI head-on: 3,400 jobs lost in 2025, 500 in the first quarter of 2026
- Self-driving cars in 2026: where do we really stand?
- AI voice cloning: the new wave of fraud exploding in 2026
- The ghost workers of AI: who really pays the price of ChatGPT
- The hidden footprint of AI: how much water and electricity does ChatGPT really consume?
- What if AI discovered that our knowledge is incomplete?
- The medicine of tomorrow: where AI can really take us
- What is a token? And why it is probably the most important concept for understanding AI
May 202631 articles
- Can we still stop AI? The open-source trap
- Why the same work is brilliant if it's human and rubbish if it's AI
- $500 million in 30 days: the company that discovered the true price of open-access AI
- The Chinese paradox: why Chinese AIs are sometimes more permissive than American ones
- Claude Opus 4.8: Anthropic accelerates, overtakes OpenAI in value and lays the groundwork for Mythos
- The hidden economy of AI: 50 million people are talking to a virtual companion
- Deepfakes: what this technology really changes in our relationship with images
- 1 year, 5 years, 10 years, 30 years: what current research tells us about the future of AI
- Will the AI bubble burst? What 2026 is already telling us
- The "unlimited freemium" trap: how AI gets you hooked before presenting the bill
- Faked AI demos: how to stop being fooled
- Adobe, Canva, CapCut in Gemini: Google locks down visual creation
- Hell Grind: the first 90-minute film entirely created by AI arrives at Cannes
- Antigravity 2.0: Google takes on Claude Code and Cursor on their own turf
- Why AI hardware isn't taking off (and why the Ray-Ban Meta still work)
- Gemini 3.5 Flash versus GPT-5.5, Opus 4.7 and Sonnet 4.6: the comparison that changes everything
- Google I/O 2026: all the announcements decoded
- The AI startups quietly outperforming while the giants collapse
- Spotify DJ Maïa: your new AI radio presenter finally speaks English
- Will the graphic designer's job disappear? What AI is really changing in 2026
- How France can still close its AI gap in business
- AI's biggest failures: the billion-dollar products that crashed and burned
- The silicon curtain: where the AI war between the US and China really stands
- Are your school textbooks written by ChatGPT? The scandal that is blowing the lid off
- OpenAI opens its cyber weapon to Europe, Anthropic refuses: the standoff begins
- The OpenAI lock-in: how ChatGPT wants to become your only piece of software
- Are we still in control? The real problem with AI control
- Is AI conscious? Why researchers no longer dare to say no
- The French paradox: European champions on ChatGPT, laggards in the enterprise
- Higgsfield MCP: generating images and videos directly from Claude
- France is missing the enterprise AI boat: here are the real numbers
April 202617 articles
- Google I/O 2026: what's being prepared for 19 May in Mountain View
- DeepSeek V4: China rolls out its open-source weapon against OpenAI and Anthropic
- OpenAI hits back hard: GPT-5.5, Images 2.0 and a super-app in one week
- "By the end of 2026, we'll laugh at today's AIs"
- Claude Opus 4.7 & Claude Design: Anthropic doubles down
- How AI works: The Fortune Teller's Paradox
- Launch of Alyriaz.com: Welcome everyone!
- Claude Code: Your terminal gets smart
- Suno v5.5: When AI sings in your own voice
- DeepSeek V4: the AI that cuts costs by a factor of 1000
- The AI War: United States vs China (and Europe?)
- Claude Mythos: the AI that terrifies hackers
- Tutorial: How to install OpenClaw (Your own assistant)
- OpenClaw: the AI that takes control of your computer
- DeepSeek V4: The ultra-fast software architect
- Tutorial: How to install the Google Antigravity IDE
- Google Antigravity: How developers program in 2026