Perplexity Labs wants to build your apps while you grab coffee

Perplexity is done just answering your quick questions.

This week, the $9B AI startup launched Perplexity Labs, its most ambitious feature yet. Instead of spitting out a paragraph with blue source citations in three seconds, Labs takes ten minutes or longer to research, crunch numbers, and build full-fledged work products from scratch.

Labs is available right now for all $20-a-month Pro subscribers across web, iOS, and Android, with native Mac and Windows desktop apps coming soon.

From quick answers to 10-minute deep work

The premise behind Labs is simple: give the AI more time and more tools, and let it handle multi-step workflows while you go grab a coffee.

Perplexity wired together its live web browsing with real-time code execution, data processing, and image generation. When you prompt Labs, it doesn't just draft text. It runs Python in the background, analyzes real-time figures, and packages the results into an Assets tab where you can download everything it built.

The standout feature is interactive mini web apps. Labs can build and deploy simple interactive tools, like a startup marketing dashboard, a fitness tracker, or a slideshow presentation, right inside the Perplexity interface without you writing a single line of code. To give users a starting point, Perplexity also rolled out a Projects Gallery packed with templates ranging from meal planning spreadsheets to pitch decks.

Why it matters

For daily users, Labs genuinely elevates the value of the $20-a-month Pro subscription. If you are an operator or solo founder who spends hours stitching together research, Google Sheets, and basic prototypes, having an agent handle the first pass is a massive time-saver.

From an industry standpoint, we think this is Perplexity's smartest tactical pivot. Competing with Google purely on who gives a cleaner search summary is a losing game in the long run. By stretching workflows to 10-minute autonomous runs, Perplexity is moving from an AI search engine toward an autonomous digital coworker.

The real test, however, will be reliability. Running extended multi-step tasks means compounding errors: if the model misinterprets a data source early on or hits a code execution snag, waiting 10 minutes for a broken spreadsheet will frustrate users fast. Labs needs to deliver bulletproof output, not just flashy gallery demos.