ionews Curated by a person, not an algorithm
AI News · Sources

Sources worth your time

The trick to keeping up isn't following more — it's following the right few. As the how-to-follow guide puts it: pick two, not ten. Below are the sources I'd actually recommend, sorted into a fast release feed (X) and channels that explain rather than sell (YouTube). Prefer a syllabus? Structured courses now live at LinkedIn Learning on the Learn page.

Where to follow

X accounts 10

The fastest way to see releases and research the day they land — if you follow a short list. The labs announce here first; the independent voices are picked for testing things honestly rather than reposting hype.

Anthropic (@AnthropicAI) free — Model releases and safety research straight from the source — a high signal-to-noise official feed.
OpenAI (@OpenAI) free — Launch news you can't skip, though it leans corporate and hype-forward — read it for the what, not the framing.
Google DeepMind (@GoogleDeepMind) free — Gemini launches and research announcements from Google's AI lab.
AI at Meta (@AIatMeta) free — Llama weight drops, papers, and benchmarks — the home of Meta's open-model releases.
xAI (@xai) free — Grok releases, usually with raw metrics rather than polished PR.
Simon Willison (@simonw) free — Hands-on, hype-free testing of new LLM and agent tools — the most trusted independent voice in the space.
Andrej Karpathy (@karpathy) free — Capability shifts explained from first principles by someone who has built the systems.
Ethan Mollick (@emollick) free — Practical AI-at-work guidance that translates fresh research into what to actually do Monday morning.
Jim Fan (@DrJimFan) free — Frontier commentary on robotics and embodied AI from an NVIDIA practitioner, not a spectator.
AK (@_akhaliq) free — A pure paper firehose (Hugging Face Daily Papers) — zero commentary, maximum signal, for when you want the raw stream.
YouTube channels 8

For understanding rather than headlines. These teach the ideas behind the tools and are chosen to explain, not to sell — channels built around the weekly-tool-roundup cycle are deliberately left off.

3Blue1Brown free — Visual intuition for the math under machine learning — the neural-network series is the gold standard, with zero hype.
Andrej Karpathy free — First-principles LLM deep dives from a builder — long, dense, and worth every minute.
AI Explained free — Measured, benchmark-driven analysis of what new models can and can't actually do.
Two Minute Papers free — Enthusiastic but paper-grounded research summaries — the excitement is real, the sourcing is honest.
Welch Labs free — Infrequent but rigorous visual explainers that go deep on a single idea.
Computerphile free — Professor-led computer-science explainers with reliable, no-nonsense AI and ML coverage.
Yannic Kilcher free — Close-reading paper breakdowns pitched at practitioners who want the details.
StatQuest free — The clearest explanations of stats and ML fundamentals anywhere — the prerequisite layer for everything else here.

← All news