Happy Friday!
Context is Queen 👑
The hot topic of the day, as it relates to your AI harnesses, AI swarms, second brain, and all that jazz, is context. Glossa is all aboard the context train too - we extract context (findings, decisions, and risks) from discovery conversations, in addition to requirements.
I thought that was pretty great, until this week someone tried Glossa on a project outside our usual sweet spot; it raised an issue that I think might apply to everyone. He uploaded a bunch of internal process docs and asked for requirements. Glossa tried to turn his processes into requirements, rather than storing them as context.
Which makes sense … but isn’t helpful.
Almost a year ago, a customer asked me for something like this (thank you Nora, brilliant as usual): could we tag a file as “background context” - something to ground the rest of the project without turning into requirements?
I think it’s finally time to build it. Two questions for you:
Do you use docs like SOPs, org charts, internal process docs, or client readouts to inform discovery? What proportion of your projects have background context like this?
How would you want Glossa to use this info? For example: should background context shape the discovery questions we suggest? Or should we flag when a client says something that contradicts their SOPs?
Thank you, as always, for helping to make Glossa better!
What We Shipped This Week & What's Coming Next
This week:
Better questions page - filter by status (Needs Review, Open, Partially Answered, Resolved, Skipped), sort, and navigate more easily to the requirement that opened the question.
Easier team invites - during setup, paste a whole list of emails at once and set owner or member for each person. We also tidied up how invitations are managed afterward.
Clearer requirement counts - Active, canceled, suggested cancel, total - so much terminology to describe your requirements. Now there are handy tooltips explaining what each count includes (thank you Aske for this idea!).
Coming next:
More UX polish - this week is just a first taste, much more to come!
Supersession automation? - last week, we released supersession, which flags when a later conversation replaces an earlier requirement. It still has a human in the loop - but is it accurate enough that you'd want it fully automatic? Let me know!
Weekend Musing
I’ve been thinking a lot about build v buy this week - it’s something that I’m hearing a lot in sales conversations, and I’m hearing it from other vendors in the space as well. I am so curious about what really goes into a build v buy decision!
I suspect it’s not just cost, timing, and value, the levers of traditional software purchasing. My current guesses:
the joy of building
the slog of finding and evaluating a vendor
the expectation that maintenance will be less expensive as models get better
the assumption that smaller problems are safer to build
What do you think? If you’re building, buying, or deciding - I would love to hear your trains of thought.
As always, don't hesitate to reach out with questions, opinions, ideas, and have a great weekend!
- Ali
