Jamie Davies

@julianlehr @linear Ah this is very sad - we’re on business+ 😔

@viralpickaxe @linear Which plan are you on? The feature is fully rolled out, but it's on Enterprise plans only.

New: Multi-level sub-teams Structure your teams in Linear with up to 5 levels of nesting to match how your organization works.

@linear can we get the ability to nest teams more than 1 deep plz 🙏🏻

Interests are actually pretty simple - they’re just pointers that the backend can expand into additional pre-loaded data. We also try to avoid duplicating context wherever possible, including skipping interest (and other context) snapshots that haven’t changed between turns. That said, we’re also working on some more advanced kinds of interests - for example, ones that point to a sandbox in the client, allowing the agent to execute client-side tools.

@viralpickaxe super interesting article! how are you guys actually managing and retrieving context for the agent to synthesize response? It sounds like when the UI registers an interest; the hydration process spins up a sandbox for the Agent to play around in?

@theopak @byteofbits I can't take any credit here, that's all thanks to the talented @attio studio team. From watching over the shoulder I think this was some sort of figma plugin!

@viralpickaxe i love the header image! was it procedurally generated with the text? (js?) i instantly thought “dang @byteofbits already started a new dithering aesthetic” (before remembering who he is and mentally connecting your posts lol) anyway! happy user here, i appreciate this write up

@abhagsain @haydenbleasel Streamdown is great - but we currently parse content on the frontend for rendering and on the backend for other purposes. We wanted a single, consistent approach across both.

@viralpickaxe Interesting read. Thanks for sharing. For the markdown parser you mentioned you had to build a custom parser because markdown formatting isn't valid until the full text. Why didn't you use something like streamdown by @haydenbleasel ?

@zenorocha We're a pretty pure TS shop across the backend, frontend is just good olde' react with our home baked data sync library!

@viralpickaxe what an amazing read. can you tell more about the tech stack you chose to build it?

@skull8888888888 Let's do it!

@viralpickaxe incredible read! we (laminar.sh) are also based in London. Would love to meet for coffee some time and talk all things agents!

I'd just like to say that Nick was in a meeting at the time and we did actually ship this in < 1 hour

@AshPrabaker @attio Thanks for your help getting us here Ash!

@attio congrats on the launch team - been a lot of people will find a lot of value here!

Introducing Ask Attio. A new way to work with your CRM. Search, update, and create across your entire CRM by asking. Surface call insights, draft follow-ups automatically, and get answers fast. Powered by Universal Context, a unified intelligence layer native to Attio. Now live: attio.xyz/ask-now

How it started... how it's going 👇🏻

I'd love to explore open sourcing the testing side of this framework - watch this space 👀. For the actual execution of agents, there are a few open source options (e.g, OpenAI’s Agent SDK). However, these general libraries often struggle to win over a custom solution building directly on niche codebase patterns (in our case specifically - async workloads)

@viralpickaxe @attio Any plans to make this open source? :D or are you aware of something OSS similar to this framework?

At @attio, we made building AI-powered product features as routine as writing to a database. No more fragile prompts - just robust, structured code. Check out my recent post on our 🆕 eng blog (+ a sneaky peek at our homegrown internal tools 👀): attio.com/engineering/bl…

@ntanis_kou @attio I can’t take credit for the UI here! (Props to @_julianherbst and @sanjazakovska!) but we can def write more about how we approach building AI-native vs AI-bolton products - watch this space :)

@viralpickaxe @attio Appreciate it! What really stands out is how you level up features with AI in a way that feels native. I guess, how you approach product + tech when building? Also the UI always looks clean✨

Most call recording tools: 💤 Record everything 🤷‍♂️ Tell you nothing 🔗 Live in a separate tool (you probably forget to check) So we took a different approach. My team just launched Call Intelligence in @attio — real-time AI, natively in your CRM. Here’s how we built it 🧵

@pablosenabre @attio Fire away 🫡🫡🫡

@viralpickaxe @attio I really liked the product when I saw it but I like even more this way of showing up. Would love to ask you some details

@ntanis_kou @attio I'd love to do more write ups! What sorts of things / features are you most interested in?

@viralpickaxe @attio Attio keeps shipping AI that actually feels useful. As an engineer, this kind of execution is 🔥. Would be great to hear more about how you and the team approach builds like this.

@alexjamesvale @attio I'm a 3.7 fanboy but I think @byteofbits and @nicolasosharp would take it out of my salary 😔

@viralpickaxe @attio best model to use for insights generation if you were unbound by cost?

Call Intelligence is just one piece of the future we’re building at @attio - a CRM that doesn’t just store data, but understands it. We’re hiring across every team - including Automations, where we’re building magical, AI-powered features like this one. 🔗 attio.com/careers And yes - we’re already working on the next thing. (it's very cool)

Ultimately, Call Intelligence isn’t just about recording calls - it’s about making conversations a native part of your CRM. When every call is structured, searchable, and linked to the right records, your CRM stops just storing data - and starts to understand it. Let’s just say… we’re cooking. 🍳