A smart aspect that people might not realize is @Replit Agent will typically search through old tasks queue before fully starting something new, even if you prompt what feels like a new idea/ask. This is good because if there is an old task in the queue, for example I'm working on a project with 400+ task history, and you forgot something you never implemented, Agent will recommend you use the previous task, or merge into the new, to save on a) time b) tokens (don't write a new plan if you don't need to!)

Updating my Replit Prompt SKILL. With all of the recent changes Replit team has been shipping I realized its getting outdated!

Been experimenting with an interesting vibe code workflow. This has helped keep historical perspective and 'team' of agents working across a project to stay on same page, while I am still in loop. 1. Design a feature 2. Design review across agents 3. Document result, feasibility etc. into Linear issue for historical record/tracking 4. Implement/build and report back results/outcomes back into Linear issue(s) for the result and tracking I've found this works well for agent to agent collaboration and human in the loop. All of this is possible through MCP for some basic automation. I typically never create/edit issues in Linear unless I really need to.

Good tips!

Cool to see the recent Google PageSpeed Insights update adding in 'Agentic Browsing' to see how well a site is optimized for that or not. Definite sign Agentic Browsing is worth optimizing for.

It’s fun seeing employees of a company you really admire logging in and using something that you built.

Grateful to have been part of Race to Revenue as a mentor to five founders over eight weeks. They did incredible work during the cohort and have kept the momentum going since. Keep building Aaron (@ahorne), Clemens & Raphael, Natalia, and Ryan (@ryanzaucha)

This is a huge leap forward for those building products and apps globally. This solves a couple of big things: speed of hosting (nearer your users is always best) and regional compliance (countries often vary on their hosting compliance needs) Big shout out to the @Replit team on this launch!

Today @Replit Agent 4 just dropped and the multitasking is a game changer. I created a PRD generator skill in 2 minutes, auto-generated 13 prioritized tasks, and let Agent run them all simultaneously with smart dependency management. What used to take 3+ hours of sequential prompting? Done in about an hour. While 12 tasks were building in the background, I was experimenting with design variations in canvas view at the same time. The task board isn't just a UI improvement, it fundamentally changes how you work with an AI agent.

Always kind of a fun vibe coding prompt. "Please scan through the app code and look for any critical bugs or issues that we should resolve." It is a blanket prompt but sometimes you find some goodies.

2 hours to write a one pager. (Yes I still write by hand sometimes - and AI helped assist too) 5 mins to take that one pager and turn it into a sweet slide deck for core team review of it. Thanks @Replit Design

Claude's built in automated Q&A poll feature to clarify details is a tasteful way to introduce user delight and time to value. Pure product goodness.

If you aren't using Skills are you even Claude-ing?

Nice dynamic chart!