An AI that thinks with you, so the thinking stays yours. Praesto captures the conversation, maps how every concept connects, asks the questions you didn't know to ask, and ships you practice that grows in your mind.
Capture is mature. Practice is mature. The unique work is in the middle, the thinking-with-you that turns one into the other. That's the layer Praesto owns.
FSRS-scheduled free-response. Socratic chat. Self-explain gates.
Flow DAG · Clever Questions · Course-outline awareness · Calibration · Cross-conversation synthesis.
On-device audio, slide OCR, speaker diarization. The raw material Layer 2 runs on.
Most note-takers stop at the transcript. Praesto turns each session into four interlocking artifacts that drive durable understanding, not passive review.
Mic, system audio, and your slides, captured at lecture cadence with on-device speaker diarization. Drop your phone screen-down and pay attention to the room.
A live Flow DAG builds itself, typed propositional edges between concepts, anchored to the slide that introduced each one. The map of how every concept connects.
Mid-lecture Clever Questions appear at concept boundaries, not what the textbook asks, but what an attentive scholar would. You write a sentence first.
Free-response cards scheduled on FSRS, the same scheduler Anki's heaviest users have made the gold standard. The understanding stays yours.
What happens between the transcript and the flashcard. The reason walking in already knowing is possible.
As the professor speaks, Praesto extracts propositions, causes, depends‑on, applies‑to, contrasts‑with, and threads them into a directed graph anchored to the slide that introduced them. Tap any node to jump to the moment in the audio. Cross-conversation: this week's lecture wires into last week's automatically.
Not the question the slide deck telegraphs. The question an attentive scholar would ask given everything you've covered this term. Generated against your course outline, not in a vacuum, so it's challenging without being out-of-scope. You answer first; Praesto critiques the gap.
Cards are free-response, not multiple choice, the format the literature shows actually moves the needle (Karpicke & Blunt 2011, ~1.5σ). The FSRS scheduler decides when each card is due. You decide when you actually got it. The kind of knowledge that holds up.
We didn't build a productivity tool. We built a tool that makes you do the work, calibrate before you confide, retrieve before you reveal, explain before you accept. The science says that's where understanding actually grows.// Schroeder 2018, g=0.72 · Karpicke & Blunt 2011, ~1.5σ · Chi 1989 · HEPI 2025
Around half of students worry AI is hollowing out their thinking. We share that worry. Here's what we've architected so we can't accidentally do it ourselves.
The whole product is built around making you do the work, retrieve, articulate, connect. That's the only thing that produces durable knowledge.
Not us, not OpenAI, not Anthropic. The opt-out is hard-coded into our pipeline, not a checkbox in settings.
Praesto critiques your sentences. It doesn't author them. An ownership tool that wrote your essays would be a contradiction.
Premium product, premium pricing, no advertising, and no incentive to sell the attention you came here to keep.
Your lectures, meetings, and thoughts are yours. No data brokers, no university analytics dashboards, not even anonymized.
We promise to make you do the work that makes you better. Subtle but important, keeps the brand honest.
Starting with segments that already pay for study and meeting tools, and value pedagogy over party tricks.
"I haven't hand-carded since October. Praesto turns lecture into practice before I close my laptop."
"Cases, classes, and committee minutes in one searchable archive. Cross-conversation flow shows me how this week's case ties to last week's."
"The self-explain gates make me write a sentence first. The tool I'd actually show my professor."
"Three months of meetings, internalized. I walk in remembering what was actually said in October."
Meeting incumbents stop at Layer 1. Study incumbents start at Layer 3. Praesto is the bridge, and the moat is everything Layer 2 does between them.
| Otter | NotebookLM | Anki | Praesto | |
|---|---|---|---|---|
| Layer 1 · captures live lectures | 30-min cap (free) | no | no | Unlimited |
| Layer 2 · maps how concepts connect | no | Shallow | no | Flow DAG, typed edges |
| Layer 2 · asks questions you'd miss | no | no | no | Clever Questions |
| Layer 2 · grounded to your syllabus | no | no | no | Course-outline aware |
| Layer 3 · spaced-repetition deck | no | no | Manual | FSRS, automatic |
| Layer 3 · self-explain gates | no | no | no | Default-on |
| Slides processed on-device | Cloud | Cloud | n/a | On-device (FERPA) |
| Free tier | 300 min/mo | Free | Free | 5 hrs/mo, full Layer 2 |
Tools that hand you the work versus tools that take it from you. We're in the upper-right.
Three architectural decisions that mean we can't accidentally do the wrong thing, even if we wanted to.
We hash, OCR, and send the resulting text. The image stays in your phone's keychain-protected enclave.
Not ours, not OpenAI's, not Anthropic's. Hard-coded into the pipeline, not a checkbox in settings.
Other speakers' voices are processed only after explicit consent for that specific session. Never globally.
Praesto is built on the opposite premise. Every Clever Question, Practice card, and chat message is gated behind your retrieval and articulation, you write the sentence first, then the tool responds. The pedagogy literature (Karpicke & Blunt 2011, Chi 1989, Schroeder 2018) is explicit about this: the work is the learning. Praesto makes you do it.
Most US universities allow audio recording with the instructor's permission. The tool is built to help you learn rather than offload, which is the framing faculty respond to. We publish a one-page faculty briefing you can hand them.
No. Your audio, transcripts, and slides are not used to train any model, ours, OpenAI's, or Anthropic's. The opt-out is hard-coded into the processing pipeline, not a setting you have to remember to toggle.
Otter is built for 30-minute business meetings; the free tier breaks for any 50-80 minute lecture. NotebookLM is excellent at static documents but cannot capture a live class and has no spaced-repetition loop. Praesto runs all three layers, capture, understanding, and consolidation, with one continuous AI doing the cognitive work between them.
Targeting month nine post-launch. iOS first because medical and graduate students skew heavily iOS, the segments where active-recall tooling matters most.