Architecture Review
Share your technical plan and Seba reviews it eng-manager-mode — traces every data flow, names the edge cases and failure modes, checks the test coverage — then returns a critique with a recommendation per issue. Billed per run.
A review that finds the landmine, not just reads the plan
Zero silent failures
Seba traces every failure mode to make it visible — to the system, the team and the user. If something can fail quietly, that’s flagged as a critical defect.
Every error has a name
Not “handle errors.” Seba names the specific failure, what triggers it, what catches it, what the user sees, and whether it’s tested — and calls out catch-all handlers.
Shadow paths and edge cases
Every data flow has a happy path plus nil input, empty input and upstream error. Every interaction has double-click, navigate-away, slow connection and stale state.
A recommendation per issue
Each finding comes with a confidence score, the concrete tradeoff and an opinionated call — so you fix the right things in the right order.
What the review checks
The same passes an experienced eng manager runs before code starts.
- Architecture — component boundaries, coupling, single points of failure
- Data flow — happy path plus nil, empty and upstream-error shadow paths
- Error handling — named failures, what catches them, catch-all code smells
- Edge cases — double-click, navigate-away, slow connection, stale state, back button
- Test coverage — failure paths tested, not just the demo path
- Observability — logs, metrics, traces, dashboards as first-class scope
- Performance — bottlenecks, scaling characteristics, N+1 query risks
- Boring by default — proven tech over innovation tokens spent carelessly
- Scope challenge — the minimum set of changes, fewer moving parts
From a technical plan to a locked-in critique in three steps
Open Seba in Telegram
One tap into @meetseba_bot. No install, no signup forms.
Share the plan
Paste your architecture, design doc or implementation plan. Describe the data flow and what’s new.
Get the critique
Seba reviews architecture, code quality, tests and performance and returns a ranked critique — each issue with confidence, tradeoff and the exact credit cost.
When to run an architecture review
Catch the design flaw before it ships to production.
- +Before starting implementation on a non-trivial feature
- +Locking in a design doc or technical plan
- +Adding a new service, job, or integration point
- +Worried a flow has untraced failure modes
- +Checking whether test coverage matches the risk
- +Sizing up whether the plan is overbuilt or under-tested
Frequently asked questions
What does the architecture review produce?+
An eng-manager-mode critique across architecture, code quality, tests and performance. Each finding has a confidence score, the concrete tradeoff and an opinionated recommendation you act on.
Does Seba write the code?+
No. This skill reviews the plan before implementation. You leave with a sharper design and a ranked list of issues to fix — building is a separate step.
What should I share?+
Whatever describes the technical plan — a design doc, an implementation outline, an architecture sketch, or a few paragraphs of how the data flows. More detail gives a sharper review.
Does it handle strategy docs too?+
It can review any plan, but it stays eng-manager-mode — data flow, edge cases, failure modes. For premises, scope and ambition, run Strategy Review first.
How much does a review cost?+
You spend credits per run, and Seba shows the exact cost before it starts. Start free with welcome credits; credits never expire.
Will it just rubber-stamp my plan?+
No. The job is to find the landmine — silent failures, untraced shadow paths, missing test coverage. If a section genuinely has no issues, Seba says so and moves on.
Lock in your architecture in one message
Share a technical plan and get an eng-manager critique with a recommendation per issue back. Pay only for the run.
No install · no signup forms