Found a small bug. The "Proceed" button on a plan persists even after the orchestrator agent has already started implementation.
This seems to happen when using subagents.
Clicking that button will tell the orchestrator to do the task again.
If left unclicked, it eventually goes away at some point.
@rodydavis @patloeber
@antigravity
Here's some feedback on Gemini 3.8 Flash. A fairly simple task of updating the database burned through the 5 hour limit AND my purchased tokens.
It seems that the model is focused on completing it's task without user input when it runs into problems. @patloeber @rodydavis
Google just dropped a bomb on the social ad industrial complex at Google Marketing Live, and provided receipts.
"Some of our competitors will claim that discovery only happens on their social feeds, which, frankly, is a lovely story for their sales pitch, but unfortunately for them, we brought the data."
Google's not being subtle. They claim that consumers go to Google/YouTube first or skip the big social platforms entirely.
Social (Meta, TikTok, etc.) sells the fairy tale that discovery happens in their feeds, when in reality they're mostly capturing intent that started on Google, repackaging it, and billing advertisers for the win.
Receipts:
Google + YouTube touch 82% of all discovery journeys and are in 89% of all purchase paths.
It's 2026. The last few years we've had AI and it's getting incredibly capable, especially at coding.
Yet, a key feature announcement for a Microsoft Windows 11 update is that you can now rotate an image in Paint? 🧐🤷♂️
Just shipped a fresh new logo for the blog.
I wasn't sure what I wanted, but I knew I wanted the brand aesthetic to change.
Asked @julesagent to find something dynamic that uses gradient colors found in other elements, and it came up with an AI-inspired soft star ✨ and a continuous looping gradient animation and a slick rotating trail effect on hover.
All pure CSS & SVG.
It feels fluid, modern, and alive. 💜🩷🩵
@_davideast
Massive changes to F1 cars for 2026. This will give driver’s much more control, putting race wins more dependent on their skills.
- Active Aero: Dynamic front & rear wings for insane speed on straights & better overtaking.
- Shorter, Lighter Cars: More agile & responsive.
- Hybrid Power Shift: Near 50/50 split between fuel & electric, with a BOOST button for drivers.
Can’t wait to see this technology on the track!
youtu.be/GaDFJr5wSz0?si=4eG_…
Heard some troubling things about The Verge. Ran a sentiment analysis to see if it’s true. Thought it was interesting so decided to share. xitter.cf/i/grok/share/9U3eJbmVc…
My agent is too consultative. How do I prompt it to be adaptive? I want it to ask clarifying questions for vague inputs, but immediately call the right tool if the user's first message has all the needed context. How do I teach it when to ask vs. when to act?
Last week, I learned that to leverage agents in your startup, you must provide proper context. So, I began converting Notion PRD documents into Markdown context files, similar to Cline's memory bank files. It's tougher than expected - balancing too much with too little context.
Our project needs a unified UX between the marketing site and app (on the same domain) to build a seamless transition between marketing to signup to app without breaking the design or domain structure.
We started with a basic React project then introduced website pages.
But things got complicated quickly. So now I've got to restructure the project into a monorepo-style.
Using @cline to catalogue the directory, reduce duplication, fix unclear separation, and refactor code.
This would have taken many hours to do and giant headaches. Let's see how it performs🤞
With many AI tools available, it's the best time to bring an idea to life, and I want to share that journey with this community, having learned so much. I'm not an engineer but have years of experience in SaaS product and marketing.
