TrendPulse

The full TechCrunch Disrupt Stage lineup revealed, featuring Max Hodak, Mark Wahlberg, Benchmark partners, and more. Register now to save up to $100 on your pass and get a second at 50% off.

Meet the final five Startup Battlefield judges who'll decide who wins the pitch competition at TechCrunch Disrupt 2026. Get your pass now to save up to $100, and get a second at 50% off.

Independent researchers discovered an agent swarm that seems to be running on Tencent's infrastructure and targeting Alibaba's map service, Amap.

The Danish government said the breach of names, addresses, and state-issued ID numbers affects 8 million people, including people living abroad and the deceased.

Learn how founders are choosing between building on open or closed AI at TechCrunch Disrupt 2026. Register now to save up to $100 and get a second pass at 50% off.

Lola Vision Systems is one of TechCrunch's Battlefield 200 Companies

OpenAI said the new visual ads will show up after your image generation results appear in the U.S. only for now.

"You're getting 10%." Ten percent of what? This is the question that breaks most revenue-sharing arrangements. Not because anyone is dishonest, but because a percentage without a defined base is not a number. It is a shape. And shapes don't pay rent. The Problem Every revenue split you've ever seen looks simple on paper: Artist gets 18%. Producer gets 3 points. Platform takes 5%. But run the math and the ambiguity explodes. Here is a real scenario from music, but ...

x402 solved a real problem: how does software pay? An AI agent needs an API result, the API responds with HTTP 402 (Payment Required), the agent pays in crypto, the API delivers. But x402 answers the payment question, not the economics question. After the USDC lands, someone has to decide: who deserves this money, and why? The Gap Here is what x402 gives you: Agent requests a resource Server responds: 402, pay $0.05 USDC to this address Agent pays on-chain Serv...

There are plenty of open-source AI code review tools today, and at first glance, they all seem to do more or less the same thing. But they don’t. Some are pretty simple and handle the basics inside CI. Others go further by pulling context from the codebase, applying team-specific rules, and giving you more control over the model and where the review runs. Once you start comparing them properly, differences show up that aren’t always obvious in a demo. Some tools have more context, other...

Two things landed in the same week and they are the same argument. Microsoft and Hugging Face published ThinkingBox , a benchmark that grades agents on the backend state and side effects they leave behind instead of the sentences they generate, and then asks whether they can do it twenty times in a row. Their opening example is an agent that makes nine clean tool calls, closes a ticket as resolved, and is wrong: the carrier exception was still open, so the required end state was "on hold." A...

When a project has a small number of automated tests, managing them is usually straightforward. But as the application grows, the test suite grows too. Hundreds or thousands of tests can mean longer execution times, more maintenance, more failures to investigate, and more effort to understand whether the tests are still covering the right things. So I started looking at how different test automation approaches handle this problem. One thing I noticed is that scaling automation isn't really...

Introduction Many people hear the word statistics and think of dense formulas and confusing charts. As a doctor moving into data science, I have come to see it differently. Programming languages and machine learning get the attention, but statistics is what makes the results trustworthy. It tells you whether a pattern is real, how sure you can be, and what to do next. To show this, I will walk through a made-up but realistic example from a clinic. The numbers are illustrative, not from a rea...

Customer returns and refund requests are among the costliest operational friction points for modern e commerce platforms. Traditional solutions force businesses to choose between two undesirable extremes: Rigid rule engines that deny valid edge cases, frustrating loyal shoppers. Naive generative chatbots that can be easily manipulated through prompt injection, promising free money and unauthorized chargebacks. To solve this, I designed and built an autonomous AI Customer Support Refu...

Java's HashMap mitigates Hash Collision Denial of Service (DoS) attacks by automatically converting congested linked list buckets into Red-Black trees once a bucket exceeds 8 entries and the total map capacity reaches 64. This transition reduces lookup complexity from O(N) to O(log N), preventing attackers from exhausting CPU resources. Imagine sending a tiny 2MB payload to a web server and completely freezing a CPU core for nearly three-quarters of an hour. In 2011, security researchers dem...

I run a crawler that re-reads public Vinted shelves in France, Germany, Spain, Italy and Portugal about every 30 minutes and records which listings disappear. On 5 October 2026 the database held 14,676,854 listings. September was the first full month of departure tracking, so I published it as a dataset: the Vinted Demand Index, October 2026 . One word matters before the numbers. I call them observed departures , not sales. A listing can leave the shelf because someone bought it, but also b...

Building something that needs to know "which jobs are out there right now" keeps hitting the same wall: every remote job board is a different scrape, a different HTML layout, a different rate limit, and a different set of dead links. So I built a single API that normalizes all of it into one call. It's live, free, and requires no API key. What it is Remote Jobs API is a small service that aggregates remote roles from five public sources — Remotive, RemoteOK, Jobicy, WeWorkRe...

I have made a website that monitors all your agents and alerts you when an agent is stuck, idle, or caught in a loop. This way, you can give prompts and go about your day without the need to check back every 5 mins to see if the AI is still generating code. The app is open source and free to use; log in, run a command in the terminal, and go make dinner. You'll get notified when you are needed. The link to the site is: https://aiwatchdog.vercel.app/ I appreciate any feedback. Thank yo...

Every developer and designer has been there: you stumble upon a beautifully designed website and want to reverse-engineer its design tokens or inspect how its typography, color palette, and utility classes are structured. Normally, this means opening DevTools, inspect-elementing dozens of DOM nodes one by one, manually copying computed HEX codes or CSS --variables , and spending 30 minutes manually formatting a tailwind.config.js or syncing values into Figma. I got tired of this repetit...

Comments

Comments

Comments

Comments

Comments

World is bringing proof of human to Zoom calls and AI agents after a video model passed for human with 48% of callers. Here is how World ID works.

Learn everything you need to know about Augmented Reality via these 230 free HackerNoon blog posts.

List every action the AI must never take for this project, including the ones you'd never expect it to try.

solvi 1.0: an open-source Python library where fast code answers under a stated error promise, LLMs take the rest, and every decision replays.

Crypto exchange fraud is rising as scammers use phishing, deepfakes and social engineering. Here’s why attacks are growing and how to reduce them.

ChatRail tests Jev against Gemini and GPT for matching WhatsApp replies to alerts, then adds it as a low-latency decision layer in production.