Aaron Vanston

Sharing because more companies will need to make their operating model explicit. AI is changing the leverage of small teams and the pace at which companies can move. The companies that handle it well will rebuild around it with intent. method.bpass.team/method

We wrote the Method because expectations need a visible home beyond founder conversations and Slack threads. They should be durable and easy to revisit. If we say we are building for builders, we need to act like builders ourselves.

In March, we shared The BuildPass Method with our whole team. It is our internal page for how we build, who we hire, and what we expect. Really, it is about a bigger question: what does it mean to build a company in the age of AI? method.bpass.team/method

@fcoury Back for me now

Limits should be recovering. If yours is back to normal let me know please.

@thsottiaux Ability to start a workspace without sending a prompt first. Would love to have the worktree freshly setup and run some actions without sending a message prior.

For those of you living inside the codex app, what should we prioritize among features, reliability or performance?

@_colemurray @greptile @dakshgup Literally had this happen to myself trying to cancel Greptile for our team. Shared this feedback (and others) over email and got the most AI response back "You're absolutely right", I'm all for using AI and being scrappy, but this is painful.

while i appreciate the quick cancellation, forcing an email to support to cancel is exactly the bad practice california business and professions code § 17602 was written to address @greptile @dakshgup sure, "it's b2b, not consumer, not applicable" still bad CX pls fix

@ctatedev I bet, I ended up hacking together some post dev scripts to listen to ports and update env vars in the stack, worked but had limitations. This genuinely solves a huge pain point as well as solving others with cookies etc, thanks for sharing this for all!

@aaronvanston It was driving me nuts 😅

Weekend project: portless Stable, named URLs for local dev → Zero config → Subdomains for monorepos → No more port roulette or cookie bleed → Agents get a reliable URL instead of guessing ports 𝚗𝚙𝚖 𝚒 -𝚐 𝚙𝚘𝚛𝚝𝚕𝚎𝚜𝚜

@rs545837 @greptile Great benchmark! Curious if you also tracked the time of review completion?

@greptile benchmark, 50 PRs with planted bugs. Each bar is recall: what fraction of bugs the tool caught. The green bar is pure graph analysis, no LLM. The rest are LLM-based review tools. Dependency graph triage alone catches more bugs than any LLM tool. Now imagine combining both.

i spent the last few hours running some benchmarks on entity-level code review. no LLM. just tree-sitter parsing and dependency graph analysis. AACR-Bench (158 PRs, the benchmark used to evaluate GPT-5.2): matched GPT-5.2's recall. a graph walk doing the same job as a frontier model. Greptile benchmark (50 PRs with planted bugs): 73% recall. Greptile got 45%. Augment got 55%. Cursor got 41%. GitHub Copilot got 34%.

@dferber90 @rauchg Perfect, thank you!

@aaronvanston @rauchg Excited for you to try it! Sharing flags across projects and monorepos is already possible with SDK Keys, see here #how-to-use-flags-of-another-project" target="_blank" rel="nofollow noopener">vercel.com/docs/flags/ver…

Today we're announcing our very own Vercel Flags. Flags are essential to "ship fast without breaking things" as engineering teams grow. But now *every* engineering team is growing, because it's you + agents. Flags help you de-risk agentic engineering: vercel.com/docs/flags/ver…

@francedot Thanks for sharing, looks great! @nsfounder might be worth checking out superset, does not have that zen mode kbr clash in turbo repos that @conductor_build does

moved from Conductor → Superset. the pace at which tools get replaced lately is wild.

Built a tiny CLI for Granola meeting notes: granola-cli 🥣 - list/search meetings - show notes - fetch transcripts - export to markdown for Obsidian / review I’m using it to turn meetings into a searchable knowledge base. github.com/aaronvanston/g…

@avalondesign_ Technically Claude coded it. I provided emotional support and permission approvals.

@aaronvanston Weird question. Are you coding this???

Needed to test my keyboard keys with a new setup… and every top Google result was basically ads. Vibe coded one instead: real-time keyboard visualizer + event log. github.com/aaronvanston/k…

@OfficialLoganK Passion as a moat

my competitive advantage is that i'm having fun

Absolutely mind boggling to stop and realise: 1) I barely code anymore, just chat with an AI. 2) It’s not slop. I’m still engineering. I don’t feel threatened. Feels like pair programming. 3) I’m enjoying it more than coding by hand. Truly wild time to be living through.