đ” Slack gets 30 New Features at Once
Impressive, but is AI introducing feature bloat? Plus: YouTube's CEO on hard product trade-offs, Google's David East demos how to use Stitch, a new design tool raises $44m to take on Figma
Hi product people đ,
This week, Slack unveiled a major new upgrade which saw it reposition itself as an âagentic operating systemâ. The release included 30+ new features, but does super-fast, AI-powered velocity risk leading to feature bloat?
Plus, Googleâs David East gives us a masterclass in how to use Stitch, how Zapierâs CEO is measuring AI fluency in 2026 and why the huge increase in the adoption of the Command Line Interface is becoming a double edged sword for product teams.
Have a great weekend!
Rich
Watch on YouTube | Follow on Substack Notes
Key reads and resources for product teams
New from the Department of Product Substack this week:
Knowledge Series - How to simulate meetings before they happen
How to build a meeting simulator that lets you rehearse highâstakes product meetings before they happen. Instead of walking in blind to a sceptical CTO, a distracted exec, or a tense pricing review, you can spin up realistic personas, feed in your agenda and real product context, and watch the discussion play out in advance.
AI Library - Release Notes Multiplier
Use this prompt to turn your release notes into tailored content thatâs relevant for specific, different audiences including your customers, internal, senior stakeholders, engineering, customer success, sales and more. You paste in the release notes and get 5 different pieces of content you can re-use to send to specific audiences. (Department of Product)
Case study - How Pinterest saved 7,000 engineering hours a month by building an MCP ecosystem
Pinterestâs MCP platform turns the Model Context Protocol from a backend standard into a real product for engineers: a centrally governed âapp storeâ of secure, cloudâhosted MCP servers that plug directly into IDEs, internal chat, and AI agents. Pinterest says it has saved thousands of engineerâhours every month.
Tools you can use â Build working internal tools with AI (no dev needed)
Wow. Softrâs AI Cobuilder lets you spin up working product tools in minutes â with a visual editor, user permissions, data connections, and hosting all baked in. No sprint planning, no backlog trade-offs.
Ship what you need without dev time: release planning, NPS tracking, customer feedback loops, public roadmaps, issue trackers.
Build it. Validate it. Hand it off. (Softr*)
Skills - Replitâs Technical PM explains how to build working prototypes
Kofi Wood argues that deletes the messy translation points between PM, design, and engineering. You still move through discovery, defining behavior, prototyping, iteration, testing, and handoff, but an AI turns your behavioral brief directly into a working prototype in the same environment as production. (Replit)
Process - Why Evals are the new PRDs
Instead of writing PRDs that say âbe helpful and accurate,â you define concrete tests that run on every commit, capture real failures from production, and turn them into targets your team can systematically hillclimb against. (Braintrust)
Interview - YouTubeâs CEO Neil Mohan on Algorithm Design, AI Slop, and Hard Product Trade-offs
What does it take to build for two billion users? YouTubeâs CEO on algorithm design, the AI slop problem, and the trade-offs behind YouTubeâs biggest product decisions. (New York Times, YouTube)
Design - How to move fluidly between Claude Code and Figma
Join Figmaâs designer advocate, Brett McMillin, and Anthropicâs technical staff member, Thariq Shihipar, as they talk through how to move fluidly between Claude Code and Figma. (X)
Zapierâs CEO has unveiled how the company measures AI fluency
Hereâs a simple tool that lets you assess your own AI fluency levels using the principles outlined in Zapierâs assessment framework.
*Sponsored by Softr who have given DoP readers 200 free credits to try
New product features and innovation this week
Googleâs design Agent Stitch continues to get new powerful improvements and this week, they added a new âenhance promptâ tool that can improve the designs that the tool produces and also help teach you design language at the same time. So, for example, it will teach you to swap abstract words for more tangible aesthetic descriptions and understand color hierarchy.
As part of the launch, Google released a handy video demo presented by their DevRel Engineer David East which talks you through how to get set up and how to get consistently premium results:
Slack gets a major upgrade with â30 new featuresâ; impressive - but is AI encouraging product teams to ship too many features?
Slack has unveiled a major makeover of its core product with over 30 new features announced this week, including:
A new desktop companion which mode means Slackbot follows you across apps on your machine, so you can highlight anything on screen and ask it to act on it.
Meeting intelligence which goes beyond transcription and auto-updates your CRM, logs actions, and closes the loop without anyone typing anything.
AI-skills that act as reusable, shareable instruction sets that standardise recurring tasks like briefs, reports, and summaries, triggering automatically when Slackbot recognises the pattern.
And as an MCP client, it can now route work to any connected agent or app Agentforce, Google Workspace, Notion, Workday, and 2,600+ Slack apps.
The announcement centres around a major evolution of Slackbot - essentially repositioning Slack from a messaging tool into what theyâre calling an âagentic operating systemâ for the enterprise. Given that companies like Linear, Notion, Cursor and others have all spent the past few months building integrations into Slack, itâs pretty hard to disagree that Slack is becoming the glue that holds many apps together.
Shipping 30 new features is certainly impressive but hardly surprising given how AI can accelerate new feature development. Other companies have upped their release velocity in recent weeks, too, with multiple updates a week now shipping from companies like Claude and Notion. Last week, Claude Codeâs head of product put it simply by saying that the cost of shipping a new experiment is âessentially zeroâ.
In the long term, though, as Googleâs AI product leader has warned, this increased velocity could lead to feature bloat. Will users appreciate the new feature release cadence and raise their expectations or will they push back against the idea of shipping new features before theyâve had a chance to fully understand how they work?
Computer use comes to Claude Code
Computer Use is now available in Claude Code. Anthropic summed up this new feature pretty well:
For product teams, some of the potential use cases for this to experiment with include:
Testing & QA - UI testing without Playwright or test harness setup; click through flows, screenshot each state, flag regressions
Debugging visual issues - Claude Code will reproduce layout bugs at specific breakpoints on demand; captures broken state then checks relevant CSS; cutting out back-and-forth with engineering
Native & simulator testing - if your team ships iOS or macOS apps, this is probably the highest-value use case to compile, launch, and tap through iOS/macOS apps in the simulator; no XCTest setup needed; flags slow screens or broken states.
Other updates you might be interested in
Cursor has unveiled Cursor 3 - the latest major version of its code editor which recenters the whole experience around running AI agents, moving away from a traditional code editor feel toward something more like an AI workspace where multiple agents can work in parallel. For non-engineers, the main feature worth knowing about is Design Mode, where you can annotate UI elements directly in the browser, point the agent at exactly what you want changed, and iterate visually.
This new design tool raised $44 million this week. Itâs called Noon and its core value proposition is to build a place where you can âdesign how a product looks and how it worksâ in one place.
With AI Agents and even CLIs upending how traditional product work gets done, it does feel as though thereâs a huge opportunity for new tools to displace the likes of Figma. Itâs currently in a closed beta but might be worth a look if youâre interested in experimenting with new design tools.
đ Product data and trends to stay informed
Rampâs CPO Geoff Charles says API requests through Command Line Interfaces are accelerating with~3.4x growth in API-connected businesses (2,200 to 7,700) from March 2025 to February 2026.
This growth started around November, roughly matching the time when Claude Code started to experience mass adoption. The growth is a double edged sword; on one hand, more API adoption means more customers but API usage means product teams donât fully understand how their product is being used. Charles argues that this data should be used to understand what developers are building with those APIs - and then ship that feature to the ~90% of users who will never use MCP or the Command Line.
Startups are continuing to augment their workforce with AI agents. One company this week confirmed that a sevenâagent AI engineering team built 10 major product features in about a month - work its human developers would have needed roughly a month per feature to complete.
Agentic AI traffic grew 7,851% year-over-year with acceleration through Q3-Q4. Agentic browsers (AI built into browsers) are growing faster than standalone agents. E-commerce (46.6%), streaming/media (28.5%), and travel (19.2%) dominate, though agents show surprising traction in Tech/SaaS (4.1%), suggesting product evaluation use cases.
The full State of AI Traffic report breaks down how AI Agents are disrupting the web and itâs worth a read.
Apps are taking longer to get approved in the Apple App Store after vibe coding helped grow the number of apps published by 55% year on year. One vibe coder reported waiting about six weeks for initial approval, with update reviews taking two days to a week.
OpenAI has surpassed $100 in annualized ad revenue from its short, six week pilots so far.
Intercomâs principal engineer says 19.7% of its pull requests are now auto-approved by a Claude Code agent - with the goal to reach 50%+ by the end of the month.
Paid subscribers get the full DoP Substack including: The Knowledge Series for sharpening your tech / AI skills, the AI Prompt and Skills library and DoP Deep dive reports for in-depth analysis to learn lessons from the worldâs top tech companies.