The catalog

Good Our migration cut p99 latency by 42%. Here's how. (Lede first.)

Bad In today's fast-paced world, systems face many demands. Teams must consider many factors. We undertook a migration. (Real news buried four paragraphs in.)

Good A piece that maintains the same register from start to finish, or shifts register only at clear narrative beats.

Bad Section 1 opens with academic-detached prose; section 2 reads like a Slack message; section 3 switches to corporate-marketing voice. No structural reason for the shifts.

Good The plan was a tightrope: one wrong step and the whole team fell.

Bad The plan was a tightrope, and we had to keep all the plates spinning while pushing the ball down the field.

Good Our cache hit ratio jumped from 71% to 94% (Grafana, week of March 12).

Bad This is the best architecture for scale. (No data, no comparison, no reasoning.)

Good If we miss this deadline, the holiday launch slips and we lose the Q4 revenue we already booked.

Bad We discussed the trade-offs and reviewed the options thoroughly. (Reader has no idea what was at stake.)

Good Specific names, dates, numbers, judgments, and a recognizable narrator stance.

Bad Abstract claims about 'organizations', 'solutions', and 'best practices' without anyone, any date, or any number.

Good After three paragraphs of theory, an example: 'Here's what this looked like for the payments team last quarter.'

Bad Six paragraphs of 'systems', 'processes', and 'frameworks' with no concrete instance.

Good She left the meeting without speaking and the door clicked behind her.

Bad She was furious and everyone could feel the tension.

Good Each paragraph picks up a thread from the previous one and advances it.

Bad Paragraph 1 about caching. Paragraph 2 about team morale. Paragraph 3 about cost. No connective tissue.

Good Voice is consistently casual or consistently formal, with deliberate shifts.

Bad 'Furthermore, the implementation leverages a robust caching layer (it's basically just Redis, lol).' — two registers smashed together.

Good The writer takes a position and sticks with it, naming real counterarguments where they bite.

Bad Each paragraph proposes a view, then immediately walks it back with 'however'. By the end the reader has no idea what the writer thinks.

Good The opening sets a hook; the closing lands a different beat (a surprise, a stake, a call to action).

Bad Opening: 'AI is changing how we write.' Closing: 'In conclusion, AI is changing how we write.'

Good A prose linter. 49 deterministic rules, 19 LLM-judged ones. Detection-only.

Bad Mark up prose before it ships. A skill, a CLI, and forty-four rules for catching AI tells and writing-quality issues.

Good Here's what we tried, what worked, and what didn't.

Bad What a fantastic topic — this is going to be such a rich and rewarding exploration of an area that truly matters.

Good The new filter saved support ~20 tickets/week.

Bad This new sort filter is fundamentally reshaping how humanity interacts with information.

Good Cutting build time from 11 min to 90 sec changed who shipped — the on-call eng now ships during the on-call.

Bad This isn't just about efficiency. It's about transformation.

Good A coined term defined precisely on first use, then used consistently and supported with evidence.

Bad This is what I call the Engagement Doom Loop. The Engagement Doom Loop happens whenever teams …

Good Argument paragraphs that build on each other and advance a position.

Bad The first thing is X. The second thing is Y. The third thing is Z. Each as a separate one-sentence paragraph.

Good Each paragraph either adds evidence, raises an objection, or moves to the next claim.

Bad Five paragraphs that each say 'X is important' with different metaphors and no new fact.