Resources
What is behind it, in the open.
The AI services behind each app, today's test results, and what we have shipped. No secrets, no jargon.
The AI services behind each app
We use the well-known ones, directly.
Each is good at something different. We test them every day and give each job to the one doing best. You never pick.
Anthropic
Careful thinking, code, and longer writing.
OpenAI
Pulling research together into one clear answer.
Google Gemini
Reading web pages and summarizing them quickly.
Groq
Very fast sorting — what kind of job is this — and linking your notes.
Perplexity
Fresh research from the live web.
DeepSeek
A challenger we test daily; wins some thinking jobs on price.
xAI
A challenger we test daily for writing.
Mistral
A challenger we test daily for short summaries.
Web search and page reading use two small services of their own. None of this needs anything from you.
Today's test
We test the AIs every day so you don't have to.
No real test has run yet — only a dry run with an offline stand-in on 2026-09-13, which we do not count. These are our starting picks; the first real test runs tonight.
Sorting the request
no runs yet- Groq llama-3.3-70b-versatilestarting pick
- Anthropic claude-haiku-4-5candidate
- Google Gemini gemini-2.5-flash-litecandidate
Code
no runs yet- Anthropic claude-sonnet-5starting pick
- OpenAI gpt-5candidate
- DeepSeek deepseek-chatcandidate
Writing
no runs yet- Anthropic claude-sonnet-5starting pick
- OpenAI gpt-5-minicandidate
- xAI grok-4-fastcandidate
Reading pages
no runs yet- Google Gemini gemini-2.5-flashstarting pick
Connecting your notes
no runs yet- Groq llama-3.3-70b-versatilestarting pick
- Google Gemini gemini-2.5-flash-litecandidate
Thinking it through
no runs yet- Anthropic claude-sonnet-5starting pick
- OpenAI gpt-5candidate
- DeepSeek deepseek-reasonercandidate
- Anthropic claude-opus-5candidate
Research
no runs yet- Perplexity sonar-prostarting pick
- Perplexity sonarcandidate
- Google Gemini gemini-2.5-flashcandidate
Summarizing
no runs yet- Google Gemini gemini-2.5-flashstarting pick
- Anthropic claude-haiku-4-5candidate
- Mistral mistral-small-latestcandidate
Writing it up
no runs yet- OpenAI gpt-5-ministarting pick
- Anthropic claude-sonnet-5candidate
- Google Gemini gemini-2.5-flashcandidate
Double-checking
no runs yet- OpenAI gpt-5-ministarting pick
- Anthropic claude-haiku-4-5candidate
Updates
What we shipped.
Built in public. Newest first.
- 2026-09-13
A public benchmark: the same job, three ways, judged blind
We run one real task three ways — a regular AI, our app, and a professional-grade stand-in — and have independent AI judges score them without knowing which is which. We publish the results as-is and never score ourselves. See the method on the Benchmark page; the first live results post once the data sources are connected.
- 2026-09-12
We now talk to each AI service directly
No middleman in between any more. Fewer hops, real prices, real speed numbers in our daily tests.
- 2026-09-12
Keys are locked in one place, on our side
The access keys for AI services live in a locked store on our servers. They never reach your browser and you never see a key field anywhere in the product.
- 2026-09-12
Answers from your own files and notes, in two steps
Your files, earlier results and notes are linked together. A question first finds the closest pieces, then follows the links one or two steps out to pull in what is connected. Every answer shows what it used.
- 2026-09-12
The money side checked to the cent
Balance is checked before a run, the cost is written down before the result comes back, and the books are matched against payments. Tested end to end, including what happens when a balance runs out.
- 2026-09-12
3 apps at launch
Morning Briefing, Competitor Crawler and Content Pipeline are live. Pick one, answer a few questions in chat, and it runs — now or on a schedule.
Want the next update by email?