AI writes fast. Fast isn't the same as useful.
Ask an AI to "write my weekly update" and you'll get something fluent: lots of progress, big plans, problems mentioned in passing. The catch: how much of it actually happened?
A summary with nothing behind it reads well and feels shaky the moment you hit send. All it takes is your manager asking, "Where's that number from?"
Yesterly builds recaps on a simpler premise: start with a real record of your work, then summarize.
From a day to a year, each with its own shape
Yesterly comes with templates for the updates people actually write:
- Daily recap: done, in progress, blockers and what's next;
- Weekly update: a short summary, highlights, progress by project, blockers and risks, next week;
- Monthly review: overview, key projects, outcomes you can point to, lessons and improvements, next month;
- Year in review: the period covered, major work streams, accomplishments with evidence, what you learned, looking ahead;
- Custom period: pick any range and add your own instructions on top.
Each period gets its own structure, so you're not forcing one format onto every situation. Every template is yours to edit.
Check your sources first. Write only what the record supports.
After picking a period, you can preview the source notes and decide what goes into the recap and what stays out. Then hand it to your AI, with extra instructions if you like: "keep it short", "highlight the collaboration", "leave out unconfirmed numbers".
That's the difference between Yesterly and one-click report generators: every recap is grounded in your record, not invented by the model.
The rule is simple: state only what the material supports, and never make up progress or metrics. AI handles the tedious part. Judgment and wording stay with you.
Keep editing. The first draft isn't the finish line.
Once the draft is ready, edit it freely: adjust the wording, add context, cut anything that isn't quite right. When you're happy, copy it into Slack, Notion or email, or export it as Markdown or plain text.
Prefer writing in your browser? Use DeepSeek (web): Yesterly prepares the source file and prompt, you create the recap in the familiar web chat, then save it back to your computer with its sources and revisions.
Recaps can be written with your own API key or an AI subscription you already have (Codex, Claude, GitHub Copilot and more). The capture model and the recap model are set up separately, so each can do what it does best.
A habit worth building
Treat a recap as a view of your timeline, not something written from scratch:
- During the day, let notes build up, and fix them when needed;
- Before reporting, open your timeline and confirm what really happened today, this week or this month;
- Pick a template, check the sources, create a draft;
- Review and edit, then export or send.
Stick with it for a few weeks and you'll notice that updates take less time yet say more, because every sentence has a record behind it.
Finish calmly, not in a rush
A calm update starts with having the material ready. Yesterly turns a busy week into a clear summary, instead of something improvised minutes before the deadline.
The free trial includes 5 recaps, so you can see what a recap with real sources feels like. Download Yesterly, or see plans when you need more.