forĀ 7Ā years as a PM, I've watched theĀ sameĀ storyĀ repeat: ambitious productĀ roadmaps vs. theĀ reality of limited devĀ resources
eventuallyĀ youĀ justĀ accept it. the featureĀ ideasĀ thatĀ neverĀ make it. theĀ "maybeĀ nextĀ quarter" thatĀ never comes
butĀ I've alsoĀ seen what'sĀ possible with high agency product teams (look at teams like Ramp)
they donāt see this invisible separation of responsibility. both the engineering team and product team share the same passion for one thing - shipping responsible product fast
soĀ I'veĀ beenĀ building something
It's called krosswalkĀ - aĀ tool that safely guides product teams to ask questions and build features in the existing codebase - NOT a prototype
askĀ questions, getĀ answers grounded in theĀ actualĀ code
generateĀ safe, reviewable, WORKING changesĀ that get reviewed by your designers and product team before your engineers review and merge
PMs canĀ finallyĀ moveĀ fromĀ "I wishĀ weĀ couldā¦" to "here's aĀ working prototype."
lookingĀ for PMs who feelĀ thisĀ painĀ and want to join the nextĀ batch of users -> https://t.co/qFJ47U1Gn4
@SherryYanJiang super cool sherry! are you letting more folks get read access to the real repo(s) too? (for ācurrentā prod functionality questions and better specs/prds)
the best software teams will be made of 100% builders, differentiated less by title and more by what they optimize for: product intuition vs. systems rigor
the first wave is already here - AI coding tools like claude code and cursor are letting non-technical teammates create real things but only the most innovate teams like ramp are adopting this āall buildersā mindset
this isnāt the destination! its just that current tooling is optimized for engineering workflows. the path from āi have an ideaā to āthis is safe to shipā is still messy.
as more of the team becomes hands-on, we need an interface that makes changes easy to preview, easy to review, and hard to mess up
https://t.co/qFJ47U2ecC is the missing interface
happy shipping ��
forĀ 7Ā years as a PM, I've watched theĀ sameĀ storyĀ repeat: ambitious productĀ roadmaps vs. theĀ reality of limited devĀ resources
eventuallyĀ youĀ justĀ accept it. the featureĀ ideasĀ thatĀ neverĀ make it. theĀ "maybeĀ nextĀ quarter" thatĀ never comes
butĀ I've alsoĀ seen what'sĀ possible with high agency product teams (look at teams like Ramp)
they donāt see this invisible separation of responsibility. both the engineering team and product team share the same passion for one thing - shipping responsible product fast
soĀ I'veĀ beenĀ building something
It's called krosswalkĀ - aĀ tool that safely guides product teams to ask questions and build features in the existing codebase - NOT a prototype
askĀ questions, getĀ answers grounded in theĀ actualĀ code
generateĀ safe, reviewable, WORKING changesĀ that get reviewed by your designers and product team before your engineers review and merge
PMs canĀ finallyĀ moveĀ fromĀ "I wishĀ weĀ couldā¦" to "here's aĀ working prototype."
lookingĀ for PMs who feelĀ thisĀ painĀ and want to join the nextĀ batch of users -> https://t.co/qFJ47U1Gn4
this is inspiring and exactly how I envision all ambitious product teams will work in the next few years!
I've been working on a tool called krosswalk in case there are PMs/designers that are interested in building like the Ramp team!
https://t.co/eCVEkNwwkb
This has single-handedly increased our ship velocity by 100x. As a PM, I can directly make code changes, spin up endpoints, change the FE ux, etc directly from my browser or slack. Inspect spins up a PR for my Eng team to review. My day has switched from ācan you do this for meā to ācan you stamp my PRā š
the single best thing you can do is constantly reset context - if youāre building a larger feature then Iāve laid out a method that works rly well here: https://t.co/IESqG0lfHG
Learn about how we use the filesystem to improve context efficiency for tools, MCP servers, skills, terminals, chat history, and more.
https://t.co/jLHccCwVnn
7/
the magic is the handoff:
cursor ask mode ā refine requirements interactively
sprint .md files ā turn plan into focused chunks
fresh claude code agents ā re-check reality + execute precisely
each tool does what itās best at
6/
claude code will usually spin up a few sub-agents to investigate whatās needed before implementing.
this is where the quality comes from:
smaller scope, sharper context, fewer bugs.