1/ 👋🏼 Quick update on what we’ve been working on: Since leaving Google in December, we built a bunch of prototypes to explore different directions but in the last 30 days we spent more time with one focused idea and went deeper to take it to a fully working... state. As a PM in meetings for most of my day, I was always switching between tabs—jumping from one app to another to get insights about my customers, the business, and what the team should be prioritizing. We wanted to simplify this flow and cut down the time spent digging through different sources to make decisions. What if you could flatten all those sources and tools into one tab? What if you could just talk to your business? Connect your business data from Salesforce, Zendesk, Slack, etc. and get insights in minutes—not hours—fully cited, with transparent agent reasoning across all sources. ▶️ See exactly how the system derives its answer (ex. the Salesforce SOQL query + custom Zendesk filters) ▶️ Multi-pass agentic flow: determines which source/s to use, takes a broad query, applies heuristics to refine, unifies data across different sources, and improves with user feedback ▶️ Reports automate what you want to track (I would’ve killed for a daily Discord report on what the NotebookLM community was saying)show more

What we’ve learned so far from building this: The good stuff ▶️ Whenever we show this to prospective users, the reaction is always “YES PLEASE”—which is exciting. It’s easy to explain, and people immediately get the value ▶️ It plays to our skill set and expertise, and it’s fun to refine and make it work really well ▶️ Given the progress we’ve made in just a month, it feels realistic to keep improving quality and shipping useful features Opportunities ▶️ It still looks and feels very v0.1—chat UX, connectors, reports—everything relies on clunky primitives for now ▶️ Query coverage across source types needs work. It’s easy to get multi-pass agents covering ~50% of query types, but getting the next 50% will likely take 3x the effort. So while it was easy to get this first part out, the next chunk of work is going to be a lot harder ▶️ We probably need a minimum viable set of connectors and sources for this to work well for a key use case—then it’s all about execution. We’re not quite there yet, but we’re getting closer.

This is not a launch yet :) Just sharing progress and thinking out loud~ The UX is intentionally still very half-baked as we explore what to put more effort into. I also sped the video up because the initial query took 33 seconds to complete. It also feels like something is missing—not totally sure what yet. Over the next few days, we’re stepping back to discuss, reflect, and figure out what’s next. If this resonates with you, I’d love to hear what you think. What would make this actually useful for you?

Excited about the latest tech for your drone product? Our NDAA-compliant, US-made flight controllers are designed to accelerate your path to market and provide a solid platform for developing your autonomous software. Check them out! #Drones #UAV #UAS #Robotics #MadeInUSA

love to see it 👏 come present in june!

YEEEEEEA!

This idea (prior to AI) is what I had imagined the folks at @intercom building when I heard the original pitch. But they wanted to be the source of truth. Now with AI, I think you can be the source of truth because of the connectors and data flowing through the product. Wide open

@intercom Yeah! I also think that 'answering ad hoc queries' has become a big part of knowledge work but isn't the most fun or interesting - just cruft. Might as well make it easy to get answers

Pretty cool idea!

Looks v cool! Early thoughts - Could you collect business-specific quirks somewhere central, gathered from many users' Q&A interactions like "no not like that, at ACME Corp we always use the CREATED_AT field, not FILED_AT for sorting budget reports"? - Could have option to flag a report for more QA from a technical member of the team or someone who's worked with the data for longer, so they can dig into details & then ✅it off as accurate (mb after adjustments) - Are you targeting startups/indies/freelancers? For larger corps, getting sign off for liberal data access & linkage & AI processing can be a total pita - Interesting to figure out how much this ends up like a fluid/easy dashboarding tool vs. more write-once questions. Listen to/observe early adopters ig? - If you can solve data linkage/deduplication across different sources in a way that massively reduces the time & technical expertise that's typically needed to get that right, that would be amazing (maybe not your focus, but if you solve it, what you can answer & operationalise becomes much more powerful) - You thinking of branching out into actions on the data as well as queries? Like "update hubspot 'last seen' attribute with any user activity from shopify" Would love to try this if you're looking for beta testers at any point. (Esp if you had slack+gmail+google drive)

hi and thank you! - Yes: something I've been working on is how to capture group knowledge while respecting privacy/scopes/permissions. Tricky to solve but there are some precedents that we can possibly adopt here - This is a good idea re: flagging a report for more QA, and I think in the interim there are two possible paths. One is to make collaboration really simple since we have a lot of social norms and structures that basically have this QA in place (people ask questions about the data), but for teams with less resources (people/time) I think even having an extra step to double check data or look into multple interpretations is a possible option - Getting access to bigco data is a PITA. Honestly we've had the most traction with smaller companies that want to move quickly and arm their people with data. In general we need to get privacy and security right, and while I think we're building with that as a P0, it's something we need to go deeper on before we can enable certain size bigcos - I think our initial idea was collaborative dashboarding but I need to think more about this. I feel like we're scratching at the surface of *something* I can't quite put my finger on - For some sources this is very possible and not too hard to do (to create some unified view of the customer across data). We have it working across Salesforce, Zendesk, Slack, and email for example - where it can see across those silos based on email. I think there is a little bit of work to create thoughtful joins where valuable (ex. email -> customer ID) but we'd have to think of the priority of that use case (probably most relevant is stuff like usage metrics if coded to a user ID) - Yeah! We're actively working on actions for agents! Thinking about how much is proactive vs. reactive - Oh yeah, that would be great! Give me a bit to unlock Google Drive. We have the rest working but give me ~a week or two to enable Drive.

Raiza - this is great, but can you please add an email integration? That is probably the data source that the most enterprise users want to search over

that's almost done! will be ready by friday