Dependency management is one of those invisible tax items on every engineering team. Someone starts a feature, another engineer is quietly touching the same service, and by the time anyone notices the conflict it's review week and the damage is done. The usual fix is process: Slack check-ins, blocker tickets, careful coordination. It works, but it takes real effort — and it still misses things.
Intent's new AI pre-flight check flips the model. Instead of asking engineers to track and communicate dependencies manually, the check does that reconnaissance automatically every time you hit "Start Work."
What Happens When You Start Work
The moment you open the Start Work dialog on a changeset, a check runs — no configuration, no tagging, nothing to set up. You watch it work in real time as it searches through your specs and codebase. When it finishes, you get a clear breakdown across three areas:
- Suggested blockers — Work that needs to ship before this changeset can safely start. The AI cross-references your specs against other changesets and the current codebase to find genuine prerequisites you may not have registered yet.
- Conflict detection — Other changesets currently
in_progressthat overlap with yours. Not vague overlap — the AI identifies the specific spec or system where your work intersects. - Open todos — A read of your unchecked items, filtering down to only the ones that are true prerequisites to starting, not tasks that belong in the work itself.
And if you've already explicitly registered blockers the old-fashioned way, those are always surfaced too — the AI adds to them, never replaces them.
The Part That Changes Everything
The conflict and blocker detection is where this gets genuinely powerful. Traditionally you had to know to look — you'd need to be aware that another engineer was working on something adjacent, check in with them, and manually add a blocked_by relation if needed.
Now you don't have to know. The AI searches in-progress changesets across your entire project, compares their specs against yours, and flags the overlaps. If it finds something, you see exactly which changeset conflicts and why. And with one click — right there in the dialog — you can add it as a formal blocker without ever leaving the screen.
This is dependency management that runs itself. The coordination that used to live in Slack threads, standups, and careful bookkeeping now happens automatically at the moment it actually matters: before work starts.
Informed, Not Blocked
One deliberate design choice: the check is never a hard gate. No matter what the AI finds, you can choose to proceed. Maybe you've already spoken to the team about a flagged conflict. Maybe a suggested blocker isn't relevant to your specific slice of work. The check gives you the full picture — reasoning included — and trusts you to make the call.
This is the right trade-off. The goal isn't to add friction; it's to eliminate the friction that comes later when conflicts surface at the worst possible time.
Coordination as a Side Effect
The bigger shift here is philosophical. Good dependency management shouldn't require discipline and overhead from every engineer on the team. It should be a natural side effect of how you work.
When Intent's pre-flight check runs, it's doing the kind of cross-team awareness check that used to require a meeting. It knows what's in flight, what your work touches, and where the risks are — because that information already lives in your specs and codebase. All it's doing is connecting the dots at the right moment.
Start work. Get the full picture. Ship without surprises.