Computation and concurrency
2 nodesExecution, interleaving, ownership, scheduling, and what it means for work to stop.
Repository-backed technical atlas
Concepts grow here through reading, conversation, and real work. The folders help us find things; the links are where the forest gets interesting.
21 concept nodes across 9 trunks. Markdown is canonical; Git keeps the revision history.
Latest · 2026-08-25Start here
If you opened Knowledge cold, start here. The point of this file is to remove the cost of deciding what to do next.
Follow Cancellation → Retries and timeouts → Idempotency.
Before reading, answer this out loud from memory:
A request times out. What, exactly, do you know happened on the other side?
Then read the three nodes quickly. Reading is the fast intake path; do not turn the first pass into a quiz.
Afterward, answer these without looking:
If an answer feels vague, follow the nearest link and repair the model. If the answer is crisp, move on.
Five minutes: read one node, close it, explain the invariant in your own words, then check the node again.
Twenty minutes: do the default three-node walk and let one answer turn into a short voice conversation with follow-up pressure.
Forty-five minutes: add one neighboring branch. Good choices are Transactions → MVCC → Replication and consistency, or Agent loops → Authority → Trust boundaries.
The current forest is thin around memory ordering, leases, backpressure, consensus, WAL/LSM storage, GPU execution, distributed inference, browser scheduling, sandboxing, and observability.
Do not expand one of those merely because it is on this list. Prefer the frontier that a conversation, interview question, bug, paper, or piece of real work makes interesting.
When a session materially changes understanding:
log/;The handoff is disposable current context. Git history keeps the previous versions.
Latest activity · 2026-08-25
21 new · 0 strengthened · 50 links
Trunks
Execution, interleaving, ownership, scheduling, and what it means for work to stop.
Systems where communication, partial failure, duplicate work, and stale knowledge become ordinary conditions.
Debugging, review, measurement, experiment design, and deciding which explanation deserves confidence.
The machinery that turns source and configuration into running programs, repeatable builds, and owned lifetimes.
Latency, throughput, tails, memory hierarchy, profiling, and finding the owner of elapsed time.
Models, inference, agents, tools, context, evaluation, and the systems that turn probabilistic outputs into useful work.
Browsers, APIs, state, interaction, product constraints, and the seams where application behavior becomes distributed.
Identity, capability, isolation, trust, and deciding who may cause which effect.
Persistence, transactions, versions, indexes, logs, and the rules governing what readers and writers observe.