标签：#ai

Circana numbers show US gamers overwhelmingly voting with their wallets.

Google continues to report big quarterly revenue, but its AI spending has skyrocketed.

A new default three-day cooldown delays version update pull requests so maintainers and security researchers can address findings in a release before it gets into your code. The post The case for a cooldown: Why Dependabot now waits before issuing version updates appeared first on The GitHub Blog .

If you've been in tech over the last year, you've probably noticed that almost every conversation eventually ends up talking about AI. LLMs, AI Agents, RAG, MCP, prompt engineering...there's something new every week. Like many of you, I've been spending time learning these technologies, experimenting with different tools, and trying to understand where everything is heading. But while learning AI, one question kept coming back to me. Whre does AI actually fit within Enterprise Architecture? Most conversations start with the model. I think they should start with the enterprise. Looking Back Over the last two decades, I've worked through several technology shifts. Physical infrastructure → Virtualization Virtualization → Cloud Cloud → Platform Engineering Automation → Everything Every transition introduced new tools, new platforms, and new buzzwords. But something interesting never changed. Successful enterprise systems still depended on the same fundamentals: Business goals Enterprise Architecture Reliable platforms Security Data Governance Operations Technology changed. Engineering principles didn't. That's one of the reasons I don't see AI as something completely separate. AI Is Just Another Enterprise Capability Today, AI is often treated like its own universe. Dedicated AI teams. AI platforms. AI roadmaps. AI strategies. That all makes sense. But I think there's a risk if we start treating AI as something that sits outside Enterprise Architecture. AI doesn't work in isolation. It needs good data. It needs infrastructure. It needs platforms. It needs security. It needs governance. It needs integration with business applications. And, most importantly, it needs to solve a real business problem. From an architect's point of view, AI isn't an island. It's another enterprise capability. Just like databases, APIs, messaging platforms, Kubernetes, and cloud services became part of our enterprise landscape, AI is becoming another capability that needs to be architected—n

If there's a place in the universe without GPUs, Nvidia is sending them there.

Etched, founded by three Harvard dropouts, has created new chips and memory components that speed up inference on any AI model -- no GPUs required, it says.

It's impossible to talk about a new Xbox game without also talking about the state of Xbox. Microsoft's gaming division is in freefall: Recent headlines are dominated by extensive layoffs, decimated studios, and confusing strategies, most of which stem from years of bad decisions and expensive acquisitions. But it wasn't always that way. Through its […]

Gemini had over 750 million monthly users in February.

Amazon is hoping the addition of games can turn Prime Video into a one-stop entertainment destination, borrowing a strategy from Netflix, which has increasingly embraced party games over the past several years.

Aggressive training techniques sharpens threat of bad behavior by leading models.

Barrels of fun.

Expedia Group has introduced STAR, an internal AI-assisted observability platform that helps engineers investigate production incidents using service telemetry and LLMs. Built with FastAPI, Datadog, Celery, Redis, and Langfuse, STAR follows structured workflows to analyze telemetry, generate root cause assessments, and support incident response while keeping engineers in the loop. By Leela Kumili

Lawmakers are preparing to introduce an "AI Kill Switch Act" that would require AI companies to shut down or throttle their systems on orders from the Department of Homeland Security, according to a report from Politico. Reps. Ted Lieu (D-CA) and Nathaniel Moran (R-TX) are expected to introduce the legislation on Thursday. The news of […]

Today on Decoder, I’m talking with Hayden Field, The Verge’s senior AI reporter, about the major trade secrets lawsuit between Apple and OpenAI and what this tells us about OpenAI’s future. By now I’m sure most Decoder listeners are familiar with Apple’s allegations in this case. The company says a number of ex-Apple employees at […]

Imagine you are the parent of an inquisitive seven-year-old who loves his family, playing baseball, and hanging out with his friends. It's the life you've always wanted. And then the worst thing that ever could happen happens: A freak accident takes your son away. But as it turns out, you don't have to live with […]

You give the AI two examples of a new task. It understands. It completes the third example correctly. It has not changed its weights. It has not been fine-tuned. It has learned from the context of the prompt alone. This is in-context learning. It is one of the most remarkable properties of large language models. But it is not learning in the human sense. It is pattern matching. It is using the examples as a template. It is not generalizing. It is adapting. This is the distinction that matters: in-context learning is not true generalization. It is a form of rapid pattern completion. The model does not update its internal knowledge. It simply uses the examples to adjust its predictions. What Is In-Context Learning? In-context learning is the ability of a model to learn from examples provided in the prompt. The Process: The prompt contains a few examples. The model uses these examples to infer the task. It applies the inferred task to a new input. The Mechanism: The model does not update its weights. It uses the examples as a template. It generates the most likely completion. A Contrarian Take: In-Context Learning Is Not Learning. It Is Pattern Completion. We call it "learning." But it is not learning in the human sense. It is pattern completion. The model is not generalizing. It is matching patterns. How Does It Work? The mechanism of in-context learning is still debated. But there are leading theories. The Pattern Completion Theory: The model has seen similar tasks during training. The examples activate the relevant patterns. The model completes the pattern. The Induction Head Theory: The model has "induction heads" that detect repeated patterns. These heads identify the relationship between examples. They apply the relationship to the new input. A Contrarian Take: The Mechanism Is Not Important. The Outcome Is. We debate the mechanism. But the outcome is what matters. The model can learn from examples. The mechanism is a technical detail. The outcome is a practical

In 1996, Bandai sold 82 million Tamagotchis. Kids carried egg-shaped plastic keychains everywhere, frantically pressing three buttons to feed, clean, and play with a pixelated blob that would literally die if you ignored it during math class. It was the first time millions of people felt genuine emotional attachment to a piece of software. 30 years later, I rebuilt that entire experience — in the browser, with TypeScript, zero dependencies, completely open source. No app store. No download. No install. Just open a tab and adopt your pet. And in the process, I learned more about state machines, game loops, and emotional design than any computer science course ever taught me. Why Build a Virtual Pet in 2025? Three reasons: 1. Nostalgia Is a Distribution Hack People share things that trigger childhood memories. It's not rational — it's emotional. A browser-based Tamagotchi hits a nerve that no todo app or dashboard ever will. When I shared an early prototype, the response wasn't "cool tech stack." It was: "OH MY GOD I used to cry when mine died in second grade" That emotional reaction is worth more than any Product Hunt launch. 2. Game State Machines Are Criminally Underrated Every tutorial teaches state machines with traffic lights or toggle buttons. Boring. Useless. Forgettable. A virtual pet has dozens of interconnected states , real-time decay, evolution paths, conditional transitions, and edge cases that force you to actually think about state architecture. After building this, implementing complex UI flows in production apps felt trivial. 3. Not Everything Needs to Be a SaaS The indie dev world is obsessed with "revenue-generating side projects." Sometimes you should build something purely because it makes people smile. The best projects are the ones you'd use even if nobody else existed. Meet Your New Pet When you open Tamagochi, you get an egg. It hatches. A tiny pixelated creature appears. It has needs. Meet them, and it thrives. Ignore them, and... well, game

Divergence escalates the wrong population: unanimous misses auto-pass Agent Determinism Illusions (Part 7) Where this fits: This part does not continue Part 13's probe-vs-prose thread. It returns to Part 6 's L2→L3 escalation rule — Dipankar's move of treating vote disagreement as the human-review signal. Alexey Spinov's follow-up comment says that signal points at the wrong population. Two experiments check whether he is right, and what to put in the tripwire instead. Part 6 drew this control flow: L2 multi-perspective votes │ unanimous ──────────► AUTO-PASS / AUTO-REJECT │ divergence (e.g. 2–1) ► L3 human The caveat was already in the text: divergence measures ambiguity; it does not fix unanimous systematic bias. Alexey's point is sharper — and it is about routing , not about another caveat paragraph. 1. Alexey's population mismatch On the Part 6 thread, Alexey Spinov wrote (paraphrased tightly): The dangerous failures are high-confidence and directional — systematic. Systematic bias is shared across prompts, not idiosyncratic (your own P3: majority voting doesn't fix it). So the three perspectives will tend to agree on exactly those cases. Divergence-to-human then routes you the safely-ambiguous ones and auto-passes the confidently-wrong ones. The escalation signal is pointing at the wrong population. He proposed two cheap replacements: T1 — deterministic tripwire on known-reversal classes (escalate regardless of agreement). T2 — treat unanimous + high-confidence on a historically reversal-prone class as escalate — the inverse of “high confidence, auto-pass.” That is the claim under test. Not “divergence is useless,” but “divergence alone is the wrong primary tripwire for the failure mode you already measured.” 2. Experiment A — offline proxy on DF v2 (no new API) Part 6's Mike Update already showed: of 96 DF v2 MISS runs, 95.8% sat at self-reported confidence ≥ 0.9 (avg 0.969). That mass is concentrated — Part 6 also reported ~80% of MISS runs from qwen3:0.5b —

In 2023, Bitly's free plan gave you 100 links per month with full analytics. In 2025, they slashed it to 5 links per month . Five. And they started showing full-screen interstitial ads before redirecting — even on links you created years ago. In 2026, their cheapest paid plan is $10/month — just to shorten links without ads. I was paying $120/year to turn long URLs into short ones. A problem that takes 3 lines of code to solve. So I solved it myself. For free. Forever. Introducing ZipLink ZipLink is a fast, self-hostable URL shortener built with Next.js and Firebase. Deploy it once, use it forever. No monthly fees. No link limits. No ads. No "upgrade to unlock analytics." Your links. Your data. Your domain. Your rules. https://yourdomain.com/abc123 → https://some-really-long-url.com/path/to/thing?utm=whatever That's it. That's the product. Except you own it completely. The State of URL Shorteners in 2026 (It's Embarrassing) Let me show you what the "industry leaders" are charging for what is essentially a database lookup: The Comparison Table Feature Bitly Rebrandly Short.io Dub.co TinyURL Pro ZipLink Monthly cost $10-$300/mo $13-$349/mo $19-$149/mo $24-$299/mo $9.99/mo $0 Annual cost $96-$2,388/yr $156-$4,188/yr $228-$1,788/yr $288-$3,588/yr $119.88/yr $0 Free tier links 5/month 10/month 1,000 total 25/month Unlimited (no analytics) Unlimited Custom domain Paid only Free (1 domain) Required Paid only Paid only Yes (yours) Click analytics Paid only Paid only Free (basic) Paid only Paid only Full, free Link expiration Paid only Paid only Yes Yes No Yes Password protection Enterprise only Paid only Paid Paid No Yes API access Paid only Paid only Yes Yes Paid Full, free QR codes 2 free, then paid Paid Yes Paid Paid Unlimited Self-hostable No No No Yes (complex) No Yes (simple) Data ownership Their servers Their servers Their servers Their servers Their servers Your Firebase Ads/interstitials Yes (free tier) No No No Yes (free tier) Never Links disappear if you stop pay

Before I hand this over, I want to tell you where I'm actually standing right now. I'm looking at the biggest mountain yet: building our hospital's own ERP and EMR — the systems that run the business side and the medical records side of the whole place. That's usually the job of specialized companies, and it's tangled up with outside regulators too. Nobody knows how long it'll take. I'm assuming at least three years. Honestly, it feels overwhelming. The size of the fear hasn't changed. My grip on it has. Here's the strange part. That overwhelmed feeling isn't unfamiliar at all. A year ago, staring at a blank screen and not knowing what code even was, I felt exactly this lost. What's different isn't the size of the fear. It's that I now know what to do with it. Get curious, look it up. Look it up, a direction shows up. See a direction, find a method, and just start. Fail, and you've learned something. Succeed, and you move to the next thing. That loop is the only thing that ever carried me anywhere , and it's the same loop I'm running at the foot of this much bigger mountain. Right now the wall isn't technical The thing I don't know isn't code. It's other departments. I moved from the treatment room to the planning office not long ago, and I still don't really know how the other departments do their work. You can't automate what you don't understand. So before anything else, I'm doing the one thing that actually comes first: listening. Hearing what people in the field actually need, what's grinding on them, and figuring out the shape of a fix together, one conversation at a time. I've noticed something before, more than once: problems that look completely different from the outside — counting a stack of forms, tracking down misplaced supplies, sorting a pile of ID cards — usually reduce to the exact same shape underneath. Make the pile of things findable. I suspect the same pattern is waiting inside every other department too. I just haven't looked yet. That's how I