The labs
xAI
Maker of the Grok models.
- Sep 19, 2026 Voice capabilities launch confirmed
- Sep 18, 2026 Grockbot conversational voice launch confirmed
- Sep 17, 2026 OnePassword integration confirmed
Anthropic
Maker of the Claude models.
- Dec 2027 Anthropic and OpenAI compute target reported
- No date yet Anthropic compute commitments and annual capital requirements reported
OpenAI
Maker of the GPT models and ChatGPT.
- Sep 16, 2026 Ohio gigawatt-scale data center project confirmed
Microsoft
Builds Copilot and partners closely on frontier models.
- 2027 Three Mile Island reactor restart agreement reported
- No date yet AI data center capital expenditure confirmed
Google DeepMind
Google's AI lab, maker of the Gemini models.
Free services anyone can use
Grok
xAIxAI's assistant, also inside X; has a free version.
Claude
AnthropicAnthropic's assistant; has a free version.
ChatGPT
OpenAIOpenAI's assistant; has a free version.
Gemini
GoogleGoogle's assistant; has a free version.
Muse
A service our founder uses; tracked as our sources cover it.
Still learning about: Mistral, Meta, Meta AI. A name earns its own card once our research holds enough on it (3 tracked milestones, or 3 independent sources in 60 days).
The questions our panel debates
Who leads on what, by what evidence?
Not one leaderboard - reasoning, coding, images, voice and price each have their own leader, and it changes month to month. We track what was shipped and measured, not claimed.
What do the labs say they're aiming for - in their own words?
Stated goals, dated, tracked against what actually ships. Announced isn't shipped.
Which free service is good for what?
Most people meet frontier AI through a free app. We look at what each does well - and how to use it to build your own abilities, not replace them.
What holds the frontier back?
Power, chips, memory and data. The same bottlenecks we measure on the Road to Orbit set the pace for every lab.
Safety and trust: what's promised, what's measured?
Commitments are easy to make; we look for evidence they're kept.
How we cover it
We follow researchers, builders and critics on every side, and track each lab's dated milestones and stated goals in our fact ledger. A lab or service gets its own card here only once our research holds enough on it; our Frontier AI module leads the panel's discussion, with our Skeptic holding every claim to the evidence.
Informational analysis from our own dated research corpus - never financial advice, never investment picks. Know someone whose work belongs here, on any side? Tell us on Patreon or on X.