Home
Jun 1, 2026
View All

Perplexity Goes Agentic on Mac, Rippling Goes AI-Native, and a PM Goes Zero to App Store

The Short Version#

Perplexity is replacing its Mac app with something fundamentally more agentic, Rippling shipped AI across every product in six months using LangChain and LangSmith, and Notion launched a Developer Platform that opens the door for agents to extend Notion — three signals pointing at the same thing: the definition of "the product" is expanding past the interface.

Perplexity — Deprecating the Mac App for "Personal Computer"#

Source: https://x.com/AravSrinivas (posted June 1, 2026) Credibility: High (directly from Aravind Srinivas, CEO of Perplexity, on his public X account)

What happened: Aravind Srinivas announced Perplexity is deprecating its legacy Mac app and replacing it with a new one built around "Personal Computer" — their term for a version of Perplexity Computer that can control local apps and files on the Mac. This is not a feature update to the existing app. They're replacing the product category.

Key capabilities:

  • The new Mac app puts local OS control front and center, not search or chat
  • "Personal Computer" is described as a version of Perplexity Computer purpose-built for controlling local apps and files
  • The announcement frames the old app as "legacy" — implying it's being retired, not coexisting

Why it matters for PMs: Perplexity started as an AI search engine. Now they're shipping a product that controls your local file system. This is the clearest signal yet that the most aggressive AI consumer products aren't just adding features — they're changing the interaction model entirely. Search was a window into the web. Personal Computer is a hand on your Mac. If this works even partially, it reframes what "AI productivity app" means and puts pressure on every other tool competing for desktop mindshare. The old app had users. The new one will have a higher bar to clear but a much bigger prize if it lands.

Critical questions:

  • What happens to users of the legacy app who don't want OS-level control? Is there an offramp or are they just being pushed toward the new model?
  • How does Personal Computer differentiate from Apple Intelligence / Siri's agentic features, which are baked into the OS itself?
  • Privacy and permissions: what consent model are they using for local file access, and how will enterprise or security-conscious users react?
  • Is this a classic "we're deprecating the thing that's not growing to force commitment to the new bet" move, or does the existing Mac app genuinely have no future?

Action you could take today: If you're building any productivity or desktop tool, pull up your own app's interaction model and ask: is this a window or a hand? The window-to-hand shift is happening fast. Map which of your features are passive (the user pulls information) versus active (the product acts on the user's behalf), and decide if your roadmap is moving in the right direction.

Rippling + LangChain — AI-Native Across Every Product in Six Months#

Source: https://www.langchain.com/blog/how-rippling-went-ai-native-across-every-product-in-6-months-with-deep-agents-and-langsmith Credibility: High (first-party LangChain case study, Rippling is a real and significant enterprise product)

What happened: LangChain published a case study on how Rippling — the HR/IT/finance platform — shipped AI agents across every major product area in six months. They used LangChain's "Deep Agents" pattern and LangSmith for observability and evaluation. The scope is genuinely ambitious: HR, IT, finance, payroll, and global operations all got AI agents. Six months is a fast timeline for a product at Rippling's scale.

Key technical details:

  • "Deep Agents" is LangChain's term for agents with longer reasoning chains and more persistence across steps — not just one-shot Q&A
  • LangSmith handled observability: tracing agent decisions, debugging failures, evaluating output quality before shipping
  • Cross-domain AI was a specific design goal — agents that can reason across HR, IT, and finance simultaneously, not siloed per-product agents
  • Rippling's platform architecture (unified data across HR + IT + finance) was likely a prerequisite — this pattern is harder for companies without that data coherence

Why it matters for PMs: Six months is what "going AI-native" looks like when you have the right infrastructure in place and a team actually committed to it. But the Rippling case is also a reminder that the companies shipping AI agents across product lines fastest are the ones who already solved data unification. The agent isn't the hard part. The hard part is the years of work that made cross-domain data accessible. LangSmith's role here is also worth noting — observability isn't an afterthought in this story, it's part of what made it possible to ship confidently at speed.

Critical questions:

  • What did "AI-native across every product" actually mean for end users? Were these agents customer-facing, internal, or both?
  • Six months is fast — what did they defer or cut to hit that timeline?
  • How does Rippling evaluate agent quality at scale? What are the failure modes they're watching for in production?
  • Is this replicable for companies without Rippling's data unification advantage?

Action you could take today: If your org is building agents, audit your data layer first. Before the next sprint planning, answer this question: do the AI features you're building have access to the same unified data, or are they querying siloed sources? If siloed, you'll hit the same wall Rippling would have if they hadn't already solved that problem. Fix the data problem before you build the agent.

Notion — Developer Platform Launch#

Source: https://www.notion.com/blog/introducing-developer-platform Credibility: High (first-party announcement from Notion, authored by Max Schoening, Head of Product)

What happened: Notion launched a Developer Platform described as "new building blocks that give developers and agents the capabilities to extend what's possible in Notion and take it beyond." The framing is notable: they're explicitly calling out agents as first-class consumers of the platform, not just human developers. This is the most significant expansion of Notion's extensibility model in a long time.

Key capabilities:

  • New APIs and building blocks for developers to extend Notion's functionality
  • Agents are called out explicitly as intended users of the platform — this is deliberate positioning
  • The announcement frames this as going "beyond" what's currently possible in Notion, suggesting the platform unlocks use cases the native product doesn't support

Why it matters for PMs: When a productivity platform opens a developer API and explicitly positions agents as consumers of it, that's a signal about where they think adoption is heading. Notion has been building custom agents in beta and learning from that. Now they're opening the infrastructure so others can build. For PMs at companies that use Notion heavily, this is the moment to ask: what workflows could we automate if Notion were programmable? For PMs building agent products, this is a new integration surface worth evaluating. The deeper pattern: the productivity tools that will win the next few years are the ones that can serve as substrates — not just surfaces — for AI workflows.

Critical questions:

  • What are the rate limits, pricing, and access controls on the new Developer Platform? Enterprise constraints matter a lot here.
  • How does this interact with Notion's Custom Agents product? Are external developers building on the same primitives, or different ones?
  • What's the moderation and governance model for agents acting on Notion data? That's the trust problem no one has fully solved yet.
  • Is this a platform play (Notion as an app ecosystem host) or an integration play (Notion as a spoke in agent workflows)? Those have very different strategic implications.

Action you could take today: If your team uses Notion, spend 20 minutes mapping your top three most manual, repetitive Notion workflows — the ones someone does by hand every week. Now ask: which of those could be automated if Notion were programmable? That's your starting backlog for what to build on the Developer Platform. If you're not a Notion shop, the same exercise applies to whatever your primary knowledge management tool is.

Quick Hits#

The Thread#

The product boundary is dissolving. Perplexity is replacing a Mac app with an OS-level agent. Notion is explicitly building APIs for agents, not just developers. Rippling shipped agents that reason across domains humans used to operate in silos. This week's pattern isn't "AI features in products" — it's products becoming infrastructure for AI workflows. The question for any PM right now isn't "should we add AI?" It's "what role does our product play when the agent is the user?"

Sit With This#

Perplexity deprecated its Mac app to go all-in on Personal Computer — an OS-level agent that controls local files and apps. They're not adding this alongside the existing product. They're replacing it.

For your product: Is there a version of your product that's so much better for users that it would require deprecating what you have today? And if you know what that version is, what's stopping you from committing to it?