Skip to content

For software engineers who want to build more ambitious things.

You can't ask for
what you can't name.

Whether you are briefing a team or directing a model, what you get depends on how precisely you can say what you want. Restudium teaches the ideas behind serious software, one at a time, until you can predict what a system will do and state exactly what it should do.

Private beta. One path of the curriculum is written.

Say it precisely1 of 4
  1. A word
  2. An arrow
  3. A guarantee
  4. Keep

One missing word.

You ask for this: “Retry failed payment calls up to three times.” The first attempt times out, although the charge went through.

What happens next?

Three short exercises. Nothing here is saved.

The language
comes first.

If you don't know an idea exists, you can't use it and you can't ask anyone for it. Precision is what turns what you know about a domain into something a team, or a model, can build from.

Without the words

“Make sync reliable.”

With them

“At-least-once delivery, an idempotency key on every write, tombstones for deletes.”

Same engineer, same tools. The second one knew what to ask for.

  1. Name it.

    Meet the idea, the problem it answers and the word for it. A name is what lets you ask for it, search for it and recognise it in someone else's design.

  2. Predict it.

    Given a concrete situation, say what the system will do before you are told. This is the depth that matters: enough to see consequences coming.

  3. Decide with it.

    Face a design choice where the idea is one of the options. Learn what it costs and where it is the wrong tool.

  4. Place it.

    Find where it already applies in a system you work on, in your own words. Ideas return later as new questions, a few at a time.

A curriculum, not a reading list.

Six areas of engineering judgment, each a set of short paths. The books and papers where these ideas were worked out are the sources, and they stay credited.

See the whole curriculum
  1. Data in more than one place

    Replication and consistency models · Partitions, and what CAP actually says · Clocks, order and causality · and 3 more

  2. Designing for failure

    Timeouts, retries and backoff · Idempotency and delivery guarantees · Backpressure and load shedding · and 2 more

  3. Saying exactly what you mean

    State machines and statecharts · Invariants, preconditions and contracts · Modelling a domain in its own words · and 2 more

  4. Knowing it is right before you ship

    Property-based testing · Deterministic simulation testing · Model checking a design · and 2 more

  5. Where data rests

    B-trees and LSM trees · Write-ahead logs and durability · Indexes, and how queries use them · and 1 more

  6. How machines talk and wait

    TCP, loss and congestion control · Event loops: epoll and io_uring · Latency, tails and queueing · and 1 more

One path is written. The rest of the map is the plan.

Confidence you have tested.

Engineers hold back from larger changes when they can't predict what will happen. Here, every answer is committed with how sure you are. Over time you see where your judgment holds and where it doesn't, and that is what lets you act on it.

Sure18 of 20 right
Fairly sure11 of 16 right
Guessing3 of 9 right

An illustrative record. Yours fills in as you answer.

New ideas at your desk.
Reviews anywhere.

At your desk

One new concept in a focused half hour, ending with where it applies in your own work.

On your phone

Up to three short reviews of ideas you have already met. Enough to keep them and keep a streak going, and capped so they never pile up.

Why I built this.

By Titus, creator of Restudium

I kept buying technical books — DDIA, SICP, the classics — reading them cover to cover, and retaining almost nothing six months later. I'd look at my bookshelf and feel like a fraud. All those ideas, just gone.

I tried Anki. Making cards by hand for a 500-page book is a second job. So the first Restudium turned a book into short lessons with a question after each. I built it carefully and used it for weeks. It felt like a chore, and I could not point to anything I could do afterwards that I couldn't do before.

What I wanted from those books was never the pages. It was the ideas: that networks partition, what consensus buys you, why a retry needs a key. Without the words, you can't ask for the right thing, from a colleague or from an AI.

I also see good engineers hold back from larger changes. Usually it isn't caution. They can't predict what will happen, so they don't trust themselves to try.

So this version teaches ideas one at a time, to the depth where you can predict and decide, and it brings each one back before you lose it. The books and papers stay credited as the sources. AI helps with preparation; deciding what is worth learning and whether a lesson works is still part of the job.

I'm building this for myself first. I want something that feels good to use, helps me make steady progress and gives the ideas a chance to stay with me. That's what Restudium means to me — study, again.

The thinking behind the practice.

Explore how recall, spaced practice and thoughtful teaching can help you build understanding that lasts.

Explore all articles

Want to know
when there is more?

Restudium is in private beta, and the curriculum is being written one path at a time. Join the list and we'll email when an invitation is available.

A few things to know.

Can I start learning today?

Access is by invitation. Joining the beta list registers your interest; it does not create an account or unlock the app. There is no announced date for open access.

What will I learn?

Six areas of engineering judgment: data in more than one place, designing for failure, saying exactly what you mean, knowing it is right before you ship, where data rests, and how machines talk and wait. Each idea is taught to the depth where you can predict and decide. That is shallower than implementing it yourself and deeper than having heard of it.

How much of the curriculum exists?

One path is written: CRDTs and local-first sync, eight concepts. The other paths on the curriculum page are planned and labelled that way.

Why not just ask an AI?

Do, once you know what to ask. A chat answers the question you bring. Restudium gives you the questions: ideas you did not know existed, in an order that builds, brought back until they stay.

Where do the paths come from?

Each path is drafted with AI assistance from primary sources: the books and papers where the ideas were worked out. Every concept lists its reading so you can check it. Generated teaching can be wrong, and review is part of preparing a path.

How does review work?

A few days after you finish a concept's questions, one new question on it comes back. Answer it and the concept counts as kept. Miss it and it returns the next day. At most three reviews are offered in a day.

What do points tell me?

Each answer is committed with how sure you are. Sure and right earns the most, sure and wrong costs a little. Points and streaks mark practice; the record of confidence against results is the part worth reading.