drew dillon

CEO & co-founder of @usebrief, AI Ships. Brief Navigates.

(it's free!)

@joinodf - ODF4
@CharacterVC - Labs G2
@speedrun - sr005

[drew dillon @drewdil]

Today we're launching Brief into open beta.

Your AI Product Navigator. The product layer for AI-powered teams.

Built with 100+ teams shipping with agents.

[drew dillon @drewdil]

I'm so bored of debates over remote work, 996, open office plans, blah, blah, blah.

these things are all tradeoffs, there's no correct answer, just a lot of signaling that amounts to personal preference for one set of tradeoffs.

[drew dillon @drewdil]

Building is cheap. agreeing on what to build is the part that still takes humans in a room.

[drew dillon @drewdil]

@thdxr GraphQL is much more useful for agents than people

[drew dillon @drewdil]

@__apf__ Posso Robulls is the one that get me

[drew dillon @drewdil]

@pxue and we call it https://t.co/68SQaXXXe2

[drew dillon @drewdil]

not coincidentally, you can find all of these primitives in @usebrief.

it's my $30k/mo consulting service for a user license fee.

[drew dillon @drewdil]

over the course of 6 years, I helped 22 teams more than double their engineering velocity.

there's no trick, in fact I wrote how to do it in a blog post I would share with prospects.

more important than ever today, when engineering velocity increasingly means token efficiency.

[drew dillon @drewdil]

real quick:
Separate Evaluation and Execution  -  slow down, consider what you’re doing

Find Yourself  -  understand your position and what will make you successful

Set Goals  - figure out what you need to get done in the next 6–9 months

Categorize  -  group strategic features by their importance against your goals

Prioritize  - stack your strategic features against burning issues and revenue opportunities

Estimate  -  balance level of effort against priority

Roadmap  - communicate themes and map the most critical features to quarters

Execute

Scale

Let It Go!

[drew dillon @drewdil]

@matt_slotnick space lasers

[drew dillon @drewdil]

.@vercel's deepsec + @OpenAI Codex /goal is a crazy combination.

GPT 5.5 chewed through improvements for days, didn't break a thing.

[drew dillon @drewdil]

it is insane that no model provider has added a "no freaking em dashes" rule. I feel like agents are over doing it just to troll me now.

hi—Drew—

[drew dillon @drewdil]

@JayaGup10 I think people see this as a software process problem, rather than a product process problem.

In Big Co. speak, PDLC not SDLC.

[drew dillon @drewdil]

Put "always use Vitest, never Jest" in the CLAUDE.md. Capitalized it. Agent skims the file, cherry-picks the section it likes, scaffolds the test in Jest anyway. The rule was right there on line 12.

[drew dillon @drewdil]

@pxue Yeah, particularly strong w/GPT 5.5, because it’s nitpicky and doesn’t give up.

[drew dillon @drewdil]

goated prompt, have run it three times already. each massive wins.

[Peter Steinberger 🦞 @steipete]

@MatthewBerman /goal refactor until you are happy with the architecture. ensure you live test after each significant step and autoreview/commit. track progress in /tmp/refactor-{projectname}.md

[drew dillon @drewdil]

code review tools apparently realized that tokens cost money over the past two weeks...

[drew dillon @drewdil]

"build me a settings page" is a spec. A bad one. Spec-driven dev just put a name on the thing you were already doing in the prompt box, usually in nine words, usually missing the part that mattered.

[drew dillon @drewdil]

@swyx I use it quite a bit, it knows to select the right cheaper models for tasks, but sometimes needs a poke to remember.

I also set up the Workflow > Codex turducken, so Claude knows to route fiddly pedantic work to GPT 5.5.