Back

examples

What a scheduled agent actually sends you

Every agent below is a real preset in Samosa, not a mockup — the title, the blurb and the schedule are pulled straight from the product. Set one going and it runs on its own, checking on the cadence shown, and delivers straight to your briefs inbox or Telegram. No tab to keep open, nothing to remember to ask for.

The output under each one is an example, written for this page to show the shape of a brief — not a real message sent to a real person.

Morning AI news

The handful of things that actually moved, before you start work.

every day at 9am

Example output
- A major model provider shipped a smaller, faster version of its flagship model — matters because inference cost, not capability, has been the thing blocking agents in production. - A widely-used open-source framework merged native support for a popular tool-calling protocol — matters because every team gluing the two together by hand can delete that code. - A large enterprise vendor said it would fold an AI assistant into its existing platform at no extra cost — matters because it resets the price a standalone competitor has to justify.

Teach me something daily

One concept a day on a topic you're learning, taught properly.

every day at 8am · What to learn: distributed systems

Example output
Today: quorum reads and writes. If a system replicates to N nodes and requires W nodes to confirm a write and R nodes to agree on a read, then W + R > N guarantees a read sees the latest write — because any read set and any write set must overlap by at least one node, and that node has the newest value. Worked example: N = 5, W = 3. What's the smallest R that still guarantees consistency? 1. Answer that first. 2. Then explain why the guarantee breaks if a node can return stale data without erroring.

Grill me on a topic

Hard questions on something you claim to know, until the gaps show.

every Wednesday at 7pm · What to be grilled on: system design

Example output
Topic: system design. Question 1 — you'd shard a database by user ID. Walk me through what happens when one user's write pattern makes their shard the hot one. Not the definition of sharding — the failure case, and what you'd actually do about it once it's already hot.

Watch a repo for releases

Know when a project you depend on ships something.

every day at 9am · GitHub repo: vercel/next.js

Example output
A new minor release landed since the last check. Two changes worth your attention: one affects the rendering path you rely on, and one breaking change moves an experimental API out of its unstable_ prefix — you'll need to update the import. The rest of the changelog is dependency bumps and internal refactors, skipped. Full notes linked.

Loose ends

The things you said you'd do and didn't.

every Friday at 5pm

Example output
- The migration doc you said you'd send the team — mentioned Tuesday, nothing since. - Following up with the vendor about the contract renewal — you said 'this week' eight days ago. - The onboarding flow you wanted a second pass on before it shipped.

What my day looks like

Your calendar, your inbox and the things you said you'd do — before the day starts.

every weekday at 7am

Example output
First up at 9:30 — the roadmap review, and it's back-to-back with the 10:15 you also have. Two unread that aren't newsletters: a reply on the contract renewal, and a calendar invite waiting on your response. Nothing in Slack is waiting on you right now. You said you'd send the revised doc before today — that's still open.

Set one of these going

Free to start. Pick a preset, change a word if you want something slightly different, and it starts working on your schedule.

Try it in chat