How to use Perplexity Computer to build a custom slack inbox (full tutorial)
TL;DR
Yash Takraal turned 100–150 daily Slack notifications into a triaged custom inbox — using Slack APIs plus AI categorization, he found that roughly 60–80% of pings were just FYIs, shrinking the real workload to about 30–40 messages.
The key move wasn’t “use AI on Slack,” it was designing a better workflow first — Takraal mapped notifications by DMs, group DMs, threads, and @mentions, then subcategorized each into action required, need to read, and FYI.
Perplexity Computer beat typical coding agents by acting like an orchestrated cloud teammate — he highlights parallel task execution, built-in connectors, and Perplexity’s model-mixing approach across Sonnet, Gemini, and Opus to fetch, plan, code, and troubleshoot with less reprompting.
The end product looks like a Kanban board for attention management — red for urgent action, yellow for read later, green for FYI, plus an “archive all” button that clears low-priority items from both the dashboard and Slack itself.
This tutorial turns into a bigger argument that SaaS isn’t dead — it’s becoming customizable — Claire Vo and Takraal argue the future is likely “Slack core + custom layers,” where niche tools can become sustainable $15/month or $10K–$20K/month businesses without venture scale.
Takraal uses the same pattern beyond Slack, from daily command centers to internal product prototyping — he built a dashboard combining AI news, email, and Slack, while a Clay teammate used Perplexity Computer to mock persona-based Clay University journeys for RevOps, SDRs, and marketers.
The Breakdown
The problem: Slack is giving him anxiety, not clarity
Yash Takraal opens with the very relatable pain point: he wakes up to 100 to 150 Slack notifications a day, many of them DMs or mentions that all feel urgent even when they’re not. The punchline is that only 30 to 40 usually matter — the rest are “FYI” noise dressed up like priority.
Before Perplexity Computer, he reverse-engineered Slack with OpenClaw
His first pass was more systems-design than magic prompt: understand Slack’s notification logic, timestamps, read states, and thread behavior, then build a deterministic digest that grouped messages into DMs, mentions, group mentions, and threads. AI only handled the fuzzy part — labeling each item as action required, need to read, or FYI — while the rest was custom code built through a long back-and-forth with OpenClaw in Discord.
The digest worked, but it still felt like scrolling through a prettier mess
The Slack digest was useful, but still too long and text-heavy to be pleasant; he says he had to scroll four or five screens just to get through it. That pushed him to a more ambitious question: what if this felt less like a log file and more like Superhuman for Slack?
Why Perplexity Computer clicked for him
Takraal says Perplexity Computer stands out because it can run long tasks in parallel, lives in the cloud, and connects natively to tools without the usual setup grind. He also loves that Perplexity is “shameless” about using the best model for each step — Sonnet for one task, Gemini for planning or Python coding, Opus for heavier reasoning — which cuts down on the frustrating reprompt loop.
The custom Slack inbox: a Kanban board for attention
The final UI is a clean board with three columns: red for action required, yellow for need to read, and green for FYIs, plus filters for DMs, group mentions, and more. The standout moment is the archive-all behavior: he can wipe the green column and those notifications disappear from both the dashboard and Slack, which made Claire immediately go, “That’s magic.”
The bigger thesis: SaaS isn’t dead, it’s getting personalized
Claire and Yash use the Slack tool to get at a bigger idea: most software is already “good enough,” but not shaped around your exact brain. Their take is that AI won’t instantly let everyone one-shot fully custom software, but it will create an explosion of small, useful apps — the kind someone would happily pay $15 a month for, or that could become a solid $10K–$20K/month niche business.
He’s building a broader personal command center, not just a Slack fix
Yash shows another app that pulls together AI news, email, and Slack into a single dashboard, again in a familiar card-based layout. He walks through the iteration honestly: first get the text right, then the UI, then usability details like deep links back to the original messages so it becomes a real command center instead of a static briefing.
His team is using the same tool for product thinking, not just productivity hacks
A teammate at Clay used Perplexity Computer to prototype persona-based Clay University flows for SDRs, BDRs, RevOps, marketing ops, and GTM engineers, all on top of the existing site. Yash frames that as the real power move: not just saving yourself time, but closing the communication gap between operators, designers, and other stakeholders by building a visual bridge everyone can react to.
The human ending: AI for board games, and yes, he threatens the model
When Claire asks about fun use cases, Yash admits he mostly treats AI as a work tool, but he does use it to brainstorm elaborate friend Olympics, organize teams for 20 people, and even discover new board games. His most memorable prompting advice is also the funniest: when AI keeps failing, he sometimes types in all caps and tells it absurdly dire consequences — “I’m going to lose my job” — and swears the extra intensity weirdly helps.