Lenss for senior engineers, tech leads and architects
Find the flaw in the RFC before the team spends the sprint on it.
Lenss audits design documents against the goal they claim to meet: the operational limit the design assumes away, the metric it reasons about wrongly, and the value at which the design stops working.
The moment this is for
Design review catches style. It rarely catches the tail-latency maths, the capacity limit or the dependency the whole proposal rests on, and those surface in production.
What it catches
Metrics reasoned about the wrong way
A design that improves the average while the success criterion is a p99, or a hit rate that cannot move the tail.
Operational limits assumed away
Capacity, throughput or cost ceilings the document states in one place and ignores in the plan.
Unstated dependencies
What the proposal needs to be true about other systems or teams, quoted where it is implied.
The breaking point
From the figures in the RFC, the value at which the design fails its own success criterion.
An example, both ways
The same draft read by a standard AI assistant and by Lenss.
Source draft
591 charsRFC: Redis cluster in front of the catalog database
Platform team · status: proposed
Adversarial review
pinned · quoted · computedPinned objective
p99 catalog read latency under 150 ms at 18,000 req/s.
Underneath: The success metric is a tail percentile; the proposal reasons about the average request.
Hidden assumption
A 90% hit rate moves the p99. It does not: p99 is set by the slowest 1% of reads, and with 10% of reads missing, that slowest 1% is all misses.
“We expect a 90% hit rate based on the 80/20 access pattern we see in logs.”
If it’s wrong: The cluster ships, average latency falls, and the SLO still fails at the same 380–450 ms tail.
Flip threshold
Cache hit rate at which p99 can fall under 150 ms
| Miss path: Postgres + cache lookup | 384–454 ms |
| Every miss vs the 150 ms SLO | over, always |
| p99 ≤ 150 ms needs misses ≤ | 1.0% of reads |
| p99 at the expected 90% hit rate | ≥ 384 ms |
At 90% the SLO fails. The decision turns on hit rate above 99%, not on the cluster.
Cheapest test: Replay one peak hour of read logs against an LRU of the planned size and measure the miss rate before provisioning.
Documents you bring
- — RFCs and design docs
- — SLA and disaster recovery specs
- — Build vs buy memos
- — Migration plans
Jobs for this role
Run it on your own draft.
Free, no account. The draft is never stored. A full audit takes two to three minutes.