WordPress took its experimental Telex thing from concept to actual production in like three months, and Matt Mullenweg showed off live business sites at State of the Word that used the AI platform to build pricing calculators and store directories through plain language prompts. The vibe-coding tool spits out Gutenberg blocks without manual coding, and one dev even cranked out a new block every day in October, including a playable ASCII Tetris game.
The company dropped an Abilities API and an MCP adapter that lets WordPress plug into AI workflows without building separate integrations for every platform, and it already works with Claude and GitHub Copilot. Real shops are using this to add dynamic features that would have cost thousands...
OpenAI grabbed Neptune, a startup that makes monitoring and debugging tools for training AI models, and the whole thing sounds like a pretty calculated infrastructure move. Neptune had been collaborating with OpenAI on metrics dashboards for foundation model teams before the acquisition went through, and the company plans to shut down external services to focus exclusively on internal OpenAI stuff.
This marks another pickup in a busy year that already saw OpenAI drop over $6 billion on Jony Ive's hardware venture, plus another $1.1 billion for Statsig. Neptune CEO Piotr Niedźwiedź basically said they built fast and precise systems for analyzing complex training workflows, which matters when you're burning months and cash on expensive...
AWS dropped point-and-click tools for building custom AI models at re:Invent, and the serverless approach through Amazon Bedrock and SageMaker AI lets developers either use a guided interface or just describe what they want in plain language. The agent-led feature handles fine-tuning automatically, and it works with Nova models plus open-source options like DeepSeek and Meta Llama.
The whole pitch centers on letting companies differentiate themselves instead of using the same off-the-shelf models as their competitors, and AWS is banking on infrastructure scale to win over enterprises that currently prefer Claude and GPT, according to surveys. Custom models mean less dependence on external AI providers while keeping tighter control over...
Google Photos just dropped its yearly recap thing, but this time, Gemini AI digs through your entire camera roll to pull out what it thinks actually mattered instead of just whatever got the most engagement. The system looks at patterns around your hobbies and relationships while analyzing photo sequences to understand context, and it can apparently tell that three trash concert pics might mean more than some perfect food shot.
The whole thing lets you hide certain people or regenerate the recap if the AI gets weird about what it picks, and there's a selfie counter that feels designed to feed social media vanity. You can export straight to CapCut for video edits or push it directly to WhatsApp Status, which basically admits that people...
Amazon dropped a new Fire TV thing where you can tell Alexa Plus to find specific movie scenes just by describing them out loud, and it actually jumps right to that moment instead of making you scrub through the whole film. The system uses Amazon Nova and Anthropic Claude to figure out what you mean when you describe character names, quotes, or whatever action is happening in the scene you want.
Right now, it only works with thousands of Prime Video movies you either bought, rented, or can watch through Prime membership, and TV show support is apparently coming later. The whole setup builds on their X-Ray feature but adds AI to understand the weird ways people actually describe scenes they barely remember.
This feels like the kind of...
Microsoft shares dropped over 2% after word got out that the company slashed sales targets for Azure AI Foundry because teams kept missing their numbers last fiscal year. The platform lets businesses build autonomous AI agents that handle tasks without humans babysitting them, but getting enterprises to actually pay for this stuff turned out way harder than expected.
The quota cut is pretty rare for Microsoft and basically admits that selling enterprise AI tools remains tough even with all the hype around the technology. Companies are still trying to figure out basic AI implementation, let alone deploying agents that run independently.
The timing sucks for Microsoft since the company has been telling everyone that AI represents its...
Kuwait just made hawala operations illegal after the Cabinet signed off on changes to the Commercial Licensing Law. The amendment targets informal money transfer networks that move cash outside official banks and exchange companies, and violators can face up to six months in jail plus fines hitting 3,000 dinars.
The government says these underground systems create a shadow economy that helps launder money and fund sketchy activities while dodging any kind of regulatory oversight. Repeat offenders or businesses caught running these operations get hit harder with potential shutdowns, asset seizures, and their court rulings published in the official gazette.
The Ministry of Commerce and Industry is pushing this as a way to lock down...
AWS just dropped something called AI Factories for governments and big companies that need to keep their data locked down while running massive AI stuff. The setup drops dedicated infrastructure straight into customer data centers and comes loaded with NVIDIA chips, Trainium processors, and all the usual AWS services like Bedrock and SageMaker.
The whole point is dodging the nightmare of building this yourself since that means buying tons of GPUs, figuring out power, and waiting years for everything to actually work. Instead, you get what's basically a private AWS region that handles compute, storage, and databases while meeting whatever compliance rules you're stuck with.
AWS and NVIDIA are working together on this, and they're...
AWS rolled out new P6e-GB300 UltraServers that pack NVIDIA GB300 NVL72 chips, and these bad boys are getting hyped as the beefiest GPU setup you can grab on Amazon EC2 right now. The hardware handles massive AI models with a trillion parameters, which basically means it can run those fancy reasoning systems at production scale.
The whole thing runs on AWS Nitro System for performance and security stuff, and it hooks up with their other services like Amazon EKS without much hassle. Companies looking to deploy huge inference workloads are probably gonna be all over this one.
