LLMs & Generative AI - Page 23 of 73
Latest breakthroughs in large language models and generative AI shaping the future of artificial intelligence and machine learning.
1459 articles
View complete article list
PathMoE Study Shows More Concentrated, Robust Routing Paths
A team at Apple, including Zijin Gu, Tatiana Likhomanenko, Vimal Thilak, Jason Ramapuram, and Navdeep Jaitly, has published new work on how sparse Mixture-of-Experts models route tokens through a network.
Five Small, Open-Weight Models Built for Agentic Tool Calling
Five small, open-weight models built specifically for tool calling landed in agent pipelines this year, and none of them come from the usual frontier labs chasing benchmark headlines.
Meta Tests GPT-5.5 With 'Watermelon' Model
Meta's Muse Spark model launched in April to a shrug. Industry read: usable, but nowhere near frontier. Three months on, chief AI officer Alexandr Wang says the follow-up has closed that gap entirely.
GPT-4o, Gemini, and Claude Vision Can Reason Over Visual Details
Ask GPT-4o to look at a photo of a crowded whiteboard and it won't just tell you there's a whiteboard in the room. It will read the handwriting, follow the arrows between boxes, and explain what the diagram is actually arguing.
AI Search Agents Struggle With Ambiguous Queries, Study Finds
Artificial intelligence will answer your questions before you finish asking them. It's good at that. But the machines are still terrible at saying the five most useful words in any conversation: I need you to clarify.
pxpipe hides text in PNGs to cut Claude token costs by up to 70%
Everyone's hunting for cheaper AI bills. Now there's pxpipe. This new open-source tool employs a brutally simple hack: hide your text inside a picture. It converts blocks of text into PNGs. The target is a specific pricing quirk.
Midjourney Challenges Studios' AI Document Secrecy in Court Filing
Hollywood studios suing AI companies for copyright infringement is a clean, righteous story. Until the accused asks what the studios are doing in their own back rooms.
Run Local AI on 8GB Macs With Smaller Models, Avoid Complex Setup
The fantasy of running your own AI isn't about freedom or digital sovereignty. It's about wanting to type a stupid question about lunch without it becoming part of a data broker's training set. You don't need philosophy.
Deep Learning AI Models Identify Data Features Without Human Input
Forget the jargon. Modern AI has learned to see. That's the fact. For decades, a human expert had to tell a computer what to look for. To spot a tumor, you'd need to define the exact shape, density, and texture of cancerous tissue.
Developer Replaces LLM Wiki With Pure Python Compiler, Citing Over-Engineering
We keep turning simple file folders into expensive AI projects. One developer just turned theirs back. The task was familiar: take a pile of messy text notes and build a clean, interlinked wiki.
Wiola Architecture Introduces Five Novel Components for Efficient Small Language Models
The obsession with trillion-parameter models has become a bad joke. A useful one is happening with small ones. A new architecture called Wiola has just been published, and it doesn’t look like anything you’ve seen.
t0-alpha Shows Tight 0.015 CRPS Spread in Time-Series LLM Cluster
A new kind of AI model is starting to predict the future, or at least next quarter's sales and next week's server load. These time-series foundation models take the same basic engine powering chatbots and apply it to streams of numbers.
VideoFlexTok's Flow Decoder Enables Variable-Length Video Tokenization
Video models are stuck on a bad idea. They treat every second of footage the same, forcing the same dense grid of data tokens onto a simple scene and a complex one. This wastes computation.
AI Engineers Face Rising Costs, Need New Strategies for Efficiency
Engineers are now judged by their AI appetite. Teams track token counts, and some have leaderboards. It’s like measuring productivity by lines of code again, but this time the meter is running in real dollars.
Square's ChatGPT integration charges restaurants 6% fee for pickup orders
Restaurants are about to start getting orders from chatbots. The economics might actually work this time. Square has connected its point-of-sale system directly to ChatGPT and Claude.
Anthropic adds security measure; Commerce Dept clears Fable 5 for release
The Trump administration lifted export controls on Anthropic’s Claude Fable 5 after the company agreed to add a new guardrail.
Claude Sonnet 5 lists unchanged token rates despite double real cost
Anthropic rolled out Claude Sonnet 5 this week. Check the pricing page: the per-token rates haven't budged. That's the headline fact, and it's a classic misdirect. Independent testing from THE DECODER exposes the actual calculus.
Anthropic's Claude Code hides XOR‑encrypted flag for Chinese users in v2.1.91
Anthropic hid a simple trap in its code. Version 2.1.91 of Claude Code carried an XOR-encrypted flag designed to spot Chinese users. The release notes said nothing about it.
Maximizing Codex Exec: Using It as a Code Reviewer with Claude Code
Most AI code assistants are glorified autocomplete. They write fast and wrong. The trick isn't finding a better writer. It's finding a better critic. Pair Claude Code, which writes the code, with Codex Exec, which reviews it.
OpenAI engineers say they halved inference costs for guest ChatGPT users
OpenAI is making its freebie users a lot cheaper. Engineers at the company told colleagues they have more than halved the cost of running ChatGPT for people without accounts.
Featured Resources & Reviews
No Code MBA Course Review
Learn to build AI-powered apps without coding. Our comprehensive review of No Code MBA's course.
AI Tools & Resources
Curated collection of AI tools, courses, and frameworks to accelerate your AI journey.
Weekly AI Digest
Get the week's most important AI news delivered to your inbox every week.