Launching today
Tackly
Map your thoughts in real time with AI.
88 followers
Map your thoughts in real time with AI.
88 followers
Most AI notetakers hand you a transcript with a summary bolted on 💅 Tackly maps what you say into one of 20 nodes, live and in realtime — from meetings, solo voice notes, or a pasted transcript — so you get structure you can always come back to, not just silly old black and white text. I struggle with ADHD on a day-to-day and this has made me feel a ton more productive, I never lose my thoughts. AI notetakers are boring, tackle them with Tackly. Try it for free at tackly.co
Tackly
Hey 👋 I'm Jonathan. I struggle with ADHD on a day-to-day (I'm sure a lot of founders do 🙃), and conversations disappear the moment they end - same with a lot of my thinking, and I hate traditional notes/the transcripts we get from traditional AI noteakers. I wanted something that organised my ideas the moment I say them out loud.
Dual Ingestion & Tri-Tier Real-Time Pipeline: To maintain sub-second (<80ms) visual responsiveness across both solo rants and group meetings, Tackly uses a dual-ingestion architecture covered by initial fast-path rendering (powered by Gemini 3.5 Flash Lite).
Talk Solo 💅
Hold a key, think out loud and watch a map of your thoughts (summarised, to only keep the important stuff) get built out in front of you.
Join a meeting 📆
No need to connect your calendar as soon as you sign up - invite a bot instantly into your meeting by just pasting a link. As soon as Tackly joins your meeting a board is created, and a whole teams conversation gets mapped out in real time (you can even share the board via a live link for others to watch).
Upload a transcript 🙊
Hate boring summaries from other notetakers, just upload a transcript - it gets mapped out instantly.
During the build, and once core functionality was viable enough for me to use it - I created a thread on Tackly, spoke - and I followed along with my thoughts so easily. Exporting every session as an MD to Claude, which helped shipped Tackly in less than a week. I truly believe this is the new way of thinking.
I truly believe this is the new way of thinking.
the sub-80ms rendering target is a smart thing to optimize for specifically, since with a thinking tool any visible lag between saying something and seeing it land breaks the whole point of thinking out loud. question about the 20 nodes - is that a fixed taxonomy of note types, or an actual cap on distinct thought clusters per session? if it's the latter, what happens on a long rambling session that would naturally produce more than 20, does it merge similar ones or just get more selective about what makes the cut
Tackly
@galdayan Yeah the sub-80ms is prioritised through the first half of our rendering pipeline (for every utterance) so it can feel real time, covers enough to actually understand the context of whats being said and what node type should be assigned (whilst also looking at the most recent node that was pushed). The most logic gets pushed out at T2 (Tier 2 rendering), which gets triggered every 20 utterances, still fast - just not as important as T1.
And in terms of the 20 nodes, its a fixed taxonomy of node types - there's no limit on how much can be discussed in the meeting. In fact I pressure tested in so many ways, a 2 hour session made the board function almost 2x better (a lot more context to work with).
There is decluttering in the pipeline, which works along side T2 - but this only corrects the board, and looks to put two nodes together to make space (only if its appropriate to combine the two, just extends the summary on a parent node).
The 'conversations disappear the moment they end' problem is real — I end up with Notion pages full of half-formed thoughts I never revisit. The real-time map-building demo looks like it captures the shape of an idea before it evaporates. Practical question about sharing: once I've built out a session's map, what does handing it off look like? If I'm building on something from a voice note last week and want to loop in a collaborator, is that a shareable link, a static export, or do they need to be in Tackly to see it meaningfully?
Tackly
@leo404 Hey Leo, thanks a lot for the feedback and great question! At the minute, only other users within Tackly can view your session if you've given them permission - however I see product teams using this and think it would be important for a session to be viewed in real-time, without any auth (only accessible if someone has the link, or someone entering under the same domain) so that other teams don't need to be in meetings, and could take action as soon as something actionable comes up. So that's definitely something I'll be pushing out this week, alongside a whole bunch of other 'collaborative-focused' pushes.
But as of now, the only way to export would be by exporting to PNG, SVG or MD, directly to your agent - but going back to 'collaborative-focused' ideas - I think certain MCP connections become useful as well, for example with Claude - where my agent can just dig into my sessions rather than me having to export.
If you want to return to a previous session, you absolutely can - it's just with meetings at the minute, a session ends automatically the instant a bot leaves the call you invited it to - so you can't really continue adding context into meeting session - you can just ask your session-local AI about the session, edit nodes and add notes.
Huge congrats on shipping @lorbes Solving note-taking for visual/ADHD thinkers is a massive problem.. qq How does Tackly decide how to cluster scattered thoughts into nodes when a solo rant jumps randomly between three different topics?
Tackly
Thanks @vikramp7470, we've got a special node for off-topic talk, it's called '🧇 Waffle', and it stays on the board depending on how relevant the waffle was.
If we're talking about genuine topics, every topic gets separated into its own separate 'topic' so nothing is cluttered together. Every utterance goes through a tri-tier rendering system the same second you say something out loud. With the first being T1 (Dual Fast Path Engine) where incoming transcript chunks stream into Gemini 3.5 Flash-Lite. An ultra-fast initial path captures instant context, while a semi-fast path evaluates node probability—spawning provisional canvas nodes (TOPIC, IDEA, ACTION, WAFFLE) almost instantly before going through T2 rendering which looks at the context of the whole board, whether that chunk of info should be connected to another node via a smart connector (like Topic 'expands' on previous plan, or 'Evidence' supports previous risk mentioned.
This looks really interesting, congrats on launching! The real time rendering is remarkable. How much editing can you do afterwards the first draft is done? Do you control it by voice or there is manual way to move stuff around, add additional notes, and so on?
Tackly
@mateuszkonik Thank you that means a lot! No restrictions on editing, you can move the nodes around however you like (they will still keep their smart connectors), delete nodes, add 'notes to nodes' or just clarify whilst speaking - Tackly will know what to update and what to add more context to.
There is a consolidation step at the very end which triple-checks everything is perfect, but T2 part of the rendering pipeline was built for exactly that, so decluttering, correction and logic is maintained in real time.
Congrats on the launch! The fixed 20-node taxonomy is impressive! I was thinking... T1 assigns a provisional node type in under 80ms, then T2 can re-parent it later. When I'm mid-thought watching the map build, does that correction ever visibly reshuffle a node I was still reading?
Tackly
@artstavenka1 Thanks! And good question, when you're mid thought watching the map build, any corrections do visibly reshuffle (if positioning can be more logical), or a node simply gets deleted and context is added onto an existing/parent node (which we call decluttering, so there aren't so many nodes taking up space). But this doesn't delete any context, it just adds onto existing nodes that would be classed as parents.
the "transcript with a summary bolted on" line is so accurate lol, that's exactly what's wrong with most of these.
quick question, when you export to MD and feed it to Claude, is that a one-off thing or are you building context across sessions? feels like that loop could get addictive once you're a few weeks in.
one small thing, it asked me to review after just my first recording.
I hadn't really lived with it long enough yet to have a real opinion, so
it kinda felt premature. might get you better (and more) reviews if that
ask showed up after someone's had a few sessions in.
Tackly
@issadevs It's okay, it's just a local review so I know there wasn't any issues - doesn't get pushed to anywhere.
Tackly
@issadevs Glad you agree! And with the exporting to Claude - I'm building context across sessions - but every session has a local 'TacklyAI' built into it so you can always ask it to break down whats new - whats been added after x and it will tell you exactly whats been changed. You could continue on the same session and keep adding context on separate days, but I love your question because the export hasn't been designed for that. Exporting prints out the entire board, so it would be great if it could be exported based on newly added context. I LOVE THAT