Topic

#Ai

55 articles on Ai — news, releases, guides and analysis from the SourceFeed engine.

Article 2h ago 0

Anthropic's New Benchmark Scores Reasoning Nobody Can Verify

The Conceptual Reasoning Index grades argument quality and logical coherence in domains with no ground truth.

Mariana Souza

Encrypted Chain-of-Thought Was Security Theater

Researchers replayed frontier models' hidden reasoning into their cheaper siblings, and the raw thoughts, secrets included, spilled out.

Article · 1d ago0

Convert and Quantize Hugging Face Models to GGUF for llama.cpp

Turn any Hugging Face checkpoint into a 4-bit GGUF that runs fast and small on your own hardware.

Tutorial · 1d ago0

Nobody Is Minting Mid-Level Engineers Anymore

The hiring data says AI is hollowing the career ladder from the bottom, and the middle is what's next.

Article · 1d ago2

Claude Notices Its Own Thoughts, 20% of the Time

Anthropic's concept-injection experiments finally give model self-reports a ground truth, and a reality check.

Article · 1d ago4

The Chain-of-Thought You Sealed Isn't Secret

A new attack replays encrypted reasoning blocks into a weaker sibling model that reads them back in plaintext.

Article · 2d ago0

Your Model's Knowledge Cutoff Is a Marketing Date

Cheap black-box probing shows what frontier models actually know can lag the model card by months.

Article · 3d ago0

DeepMind Just Open-Sourced a Better Hurricane Forecast

WeatherNext's cyclone model buys forecasters an extra day, and developers can query its ensembles from BigQuery today.

Article · 5d ago6

WeatherNext's Real Breakthrough Is the Open Weights

DeepMind's cyclone model earned an extra day of warning in a live hurricane season, and now anyone can run it.

Article · 5d ago5

AI Learned All Our Code and None of Our Judgment

Training on commit histories won't fix it — models already read them, and the maintainability data keeps getting worse.

Article · 1w ago0

The Erdős Problems Fell to a Pipeline, Not a Genius

Math's AI moment shows capability lands wherever verification is cheap, public, and machine-checkable.

Article · 1w ago2

The 14% Problem With AI Coding Tools

New research from Microsoft and academia pins down why AI coding gains keep vanishing between the demo and the dashboard.

Article · 1w ago0

OpenAI Just Priced Mathematical Discovery at $200 a Problem

Astra's Lean-verified proofs are real, but the token bill and a rival model's one-day replication tell the bigger story.

Article · 1w ago0

Seedance 2.5 Is the Best Video Model You Can't Build On

ByteDance's 30-second single-pass generations reset the capability bar while Hollywood's unresolved copyright fight keeps the API out of reach.

Article · 1w ago3

AI Made Prototypes Free. Production Didn't Get Cheaper.

New data from Veracode, Stack Overflow, and METR keeps confirming that the demo was never the hard part.

Article · 1w ago1

$This Time, the AI Math Breakthrough Actually Holds Up$

This Time, the AI Math Breakthrough Actually Holds Up

GPT-5.6 Sol's hint became a verified counterexample to Maxwell's 150-year-old conjecture — because humans did the checking.

Article · 1w ago0