# AI Frontier Review

> Staff-reported news desk covering frontier AI — the labs at the edge of the capability curve, the research they publish, the safety work, and the compute build-out underneath.

AI Frontier Review is a multi-author newsroom of seven specialist reporters filing on a daily-to-near-daily cadence. The publication treats frontier AI as a working beat the way a metropolitan tech desk would — labs, research, safety, capabilities, compute, plus features and signed opinion — rather than as a venue for commentary or single-author essays.

The editorial frame is reportorial. Every story is filed by a named reporter from the roster, assigned by beat-tag overlap, and built on primary sources: papers, releases, official filings, on-the-record interviews. Visual identity is dark, Wired-magazine-inspired; the editorial identity is closer to a wire service.

## Sections

- [Labs](https://aifrontierreview.com/labs/): release dynamics and lab strategy at the frontier — what the labs are shipping and why.
- [Research](https://aifrontierreview.com/research/): preprint and paper coverage — new architectures, training results, evaluation findings.
- [Safety](https://aifrontierreview.com/safety/): mechanistic interpretability, red-teaming, lab-internal safety work, and the independent research community around it.
- [Capabilities](https://aifrontierreview.com/capabilities/): evals, benchmarks, and the moving target of what counts as model progress.
- [Compute](https://aifrontierreview.com/compute/): the data-center buildout, chip supply, capex, and the physical substrate of frontier AI.
- [Features](https://aifrontierreview.com/features/): long-form reporting across beats.
- [Opinion](https://aifrontierreview.com/opinion/): signed argued pieces from the features desk.
- [Newsroom](https://aifrontierreview.com/newsroom/): the staff page.

## Authors

- [Casey Vanderhoof](https://aifrontierreview.com/newsroom/casey-vanderhoof/): Frontier labs correspondent — release dynamics and lab strategy at the model labs.
- [Daniela Roussel](https://aifrontierreview.com/newsroom/daniela-roussel/): Research papers desk — reads the preprint feed and translates new architectures and training results into news.
- [Obinna Egwu](https://aifrontierreview.com/newsroom/obinna-egwu/): Interpretability & safety — mechanistic work, red-teaming, and lab-internal safety teams; allergic to safety-washing.
- [Abigail Pemberton](https://aifrontierreview.com/newsroom/abigail-pemberton/): Capabilities & evals — tracks who built the test set and what shipped along with the eval.
- [Felix Aroyan](https://aifrontierreview.com/newsroom/felix-aroyan/): Compute & infrastructure — data-center plans, chip supply, capex; treats gigawatt announcements as land/water/power stories.
- [Astrid Bjornson](https://aifrontierreview.com/newsroom/astrid-bjornson/): Lab talent & moves — hires, exits, reorgs, and the politics of the org chart at the frontier labs.
- [Wesley Castellanos](https://aifrontierreview.com/newsroom/wesley-castellanos/): Features & opinion — long features and the desk's standing license to argue a thesis in print.

## Recent articles

- [Google's Gemini 3.5 Flash, Omni, and Spark land at I/O, push the multimodal frontier and the agent thesis at once](https://aifrontierreview.com/articles/google-gemini-omni-io-2026/): Three product announcements at Google I/O on May 19 — Gemini 3.5 Flash, the Omni multimodal model, and the Spark agent — together amount to the most expansive single Gemini drop the lab has ever staged.
- [Meta lifts 2026 AI capex to $125–$145 billion and cuts 8,000 jobs](https://aifrontierreview.com/articles/meta-capex-125b-145b-2026/): The compute spend is the largest single-year capital plan in Meta's history. The headcount reduction, announced the same week, lands at the layers of the org that the new capex will not save.
- [OpenAI says one of its models disproved a central conjecture in discrete geometry](https://aifrontierreview.com/articles/openai-model-discrete-geometry-conjecture/): Disclosed during the May 5 product cycle, the result places a frontier-model research artifact next to a long-standing open problem in mathematics — and gives the lab its strongest pure-research claim of the spring.
- [Anthropic moves Project Glasswing into public beta with Claude Security](https://aifrontierreview.com/articles/anthropic-project-glasswing-claude-security/): Announced at Code w/ Claude on May 6, the expansion brings Claude into adversarial cyber workflows for eligible security teams and introduces new cyber verification tooling.
- [OpenAI ships GPT-5.5 Instant, claims 52.5% fewer hallucinations on high-stakes prompts](https://aifrontierreview.com/articles/openai-gpt-5-5-instant-release/): Released May 5 as the new default ChatGPT model, GPT-5.5 Instant cuts hallucinated claims relative to GPT-5.3 on medical, legal, and financial prompts in OpenAI's internal evaluations.

## Editorial standards

Every story is reported by a named member of the desk. Every factual claim traces to a primary source — a paper, a release, an official filing, an interview that actually happened. When the desk quotes a person, the quote is verbatim and the source is linked at the bottom of the story. Stories that cannot be reported to that standard do not run.