The most excellent @dt_emmy has done it again, this time with new drag 'n' drop support into dartantic_chat, including drop target previews as you go! The dartantic_chat provides a full-featured agentic chat widget with support for all dartantic providers: pub.dev/packages/dar...

Puppeteer-style agentic control of your Flutter apps: github.com/leancodepl/m... This is huge! This provides a tight loop between the app and the agent to test features, test look and feel and fix it all while you're watching reruns of Green Acres (if you know, you know).

If only agents are reading ads in the future, what happens when ads become prompt injection attacks?

Apparently opus 4.5 is a go_router fan : )

"Wisdom is looking at what everyone else is looking at and seeing what nobody else sees." -- Neil Degrasse Tyson

ME: please check again for consistency, completeness and 100% safety of existing data even if the backup goes terribly terribly wrong during a godzilla tornado monsoon fire vulcano meteor shower storm blaze gone completely out of control. big foot. hangnail. CC: I appreciate the thoroughness!

Don't let your old ways stop you from new solutions: www.youtube.com/watch?v=B1CQ...

Today, a most excellent guest blog post by Emmanuel Tuksa on enabling image paste to the Flutter AI Toolkit (PR pending) and to the new dartantic_chat package, a fork of the AI Toolkit's chat widget simplified for dartantic. Enjoy! sellsbrothers.com/guest-post-e...

Do you have a AI-related project to share? I've built a little community for folks that want to share their AI projects, find collaborators and see what other folks are doing. Post your projects today: aiprojectboard.

I gotta say, group chats in ChatGPT provide a pretty good "let me AI that for you" experience : )

Claude Chat has some amount of compression for longer conversations, but eventually just gives up with an error message. Rude! So I used the Claude Exporter Chrome extension to download the conversation in markdown format (< 600k), fed it into a new Claude conversation and kept right on going. Nice!

I stumbled onto an interesting prompting technique today. I asked a question about task manager mcp servers and got a lot of projects that could be wrapped in mcp servers but didn't have one themselves. Frustrating. So I pushed back, clarified and got a great list back. Nothing special there.

I’m fallen immediately in love with the LM Council: lmcouncil.ai. The UX is lovely and the results are fascinating. #recommended

I'm a pragmatist and apparently so is gpt-5.2: "Philosophy is only useful if it eventually produces a config file." Truth.

My 2025 in review by ChatGPT. I wear this badge with pride: )

Lovable built a 1978 Space Invaders implementation from a single prompt that I can't tell from the original. one-shot-space-invaders.lovable.app

Got myself a Dell UltraSharp U4025QW: 40″ 5K2K (5120×2160) IPS Black @ 120 Hz, ~450 nits, Thunderbolt 4 single-cable docking, 140 W power, text so sharp it judges you. They will be prying it out of my cold dead fingers.

Looking for a hand-drawn design language for Flutter that also supports go_router or any other Navigation 2.0 based routing package? Then look no further than the latest version of Sketchy with another great update from @blaqshyd! pub.dev/packages/ske...

OH: "watching your llm do tool calls is like watching the washing machine spin. you can let it just handle it." yes you can, but you have to check your clothes for damage when they come out or you're gonna look pretty terrible when you go out on a Saturday night. : )

Excellent talk on the Amp coding agent and "leaning into the weird": youtu.be/gvIAkmZUEZY?... Amp is the coding agent that taught me the value of turning the model quality dial to 11 in spite of the increased latency and cost. #recommended

by popular demand, I've update dartantic_ai to support image editing across Google (Nano Banana!), OpenAI and Anthropic across both surfaces: API: docs.dartantic.ai/media-genera... CLI: docs.dartantic.ai/dartantic-cl... e.g. "colorize @robot_bw.png"

this is my first experience with gpt5.2-codex from within the Codex plugin for VSC. not a good start...

Instructive description of the potential of vibe engineering: simonwillison.net/2025/Dec/14/...

If you think of "primitives" as data and operations to take that data between valid states, then apps are just human-oriented UIs over primitives that no longer slow down agents because they don't have to be blocked by the UIs; agents can go directly to the primitives. youtu.be/4Bg0Q1enwS4?...

what is up with california? I'm in santa barbara at a coffee shop without wifi and with all of the doors wide open. it's 54 degrees outside! in portland, we have the sense to close the doors when it gets cold.

I wanted to be able to play with all of the dartantic functionality interactively, so I built the dartantic CLI (aren't all the cool kids building agentic CLIs these days? : ).

I find this to be an important debate presented well: youtu.be/ShsuZT_vuvU?...

On to ship #3 ⛴️: Dartantic 2.0, the Nano Banana edition! 🍌 - 12K+ LOC of changes - Unified thinking mode - Server-side tools - Media generation (images and more!) And yes: Nano Banana + Gemini 3 Pro Preview. Click here for the deets: sellsbrothers.com/dartantic-2-...

Ship #2 🛳️! This one is a big thank you to @blaqshyd for his sketchy PR that provides a circular progress indicator and a new Widget Request template: github.com/csells/sketc... so don't be shy. He's already hard at working adding (ironically) go_router support. Sketchy v0.3: pub.dev/packages/ske...

Hello and welcome to Ship Day! 🏴‍☠️ Ship #1: 🚢 We're going to start off with a PR that landed today adding macOS and web support to Lokesh's excellent rar package: pub.dev/packages/rar. I also added an example Flutter app so I could see it in action. Try it out!

Tomorrow is ship day!

Pablo Picasso said: 'The meaning of life is to find your gift. The purpose of life is to give it away,' I think he meant OSS.

hello and welcome to Sketchy v0.2! It's got at least one Linux user now (thanks for the bug fix, Daniel!), so things are getting hot in here! This time, I've implemented some more widgets suitable for implementing a chat app. Enjoy!

My process when coding with AI agents: Spec: Work w/ the agent to create a plan. Make: Agent writes the tests and the code. Inspect: Check the test results, the static analysis output and the app's features and UX. Tune: Fix the P0 issues. Hope: that your stakeholders appreciate your hard work!

I tried @elevenlabsio streaming text-to-speech API this morning prepared for a big lag between pressing the Play button and hearing anything from my laptop speakers. Boy was I wrong. I couldn't count all the way to "1" before I heard Laura talking to me. Amazing! #Recommended

This is the uncompromising successor to provider that I've been looking for: a pleasant developer experience, a simple place to start and an excellent tutorial. I'll be moving to riverpod 3 for my future Flutter projects. Thanks, Remi! riverpod.dev/docs/tutoria...

Right now I'm sitting on the couch with my phone designing an "Agent Board" to control the set of coding agents on the computer I gave them to play on: an M4 mini on the Internet so I can manage my coding agents when I'm sitting on the couch with my phone.

ChatGPT nerd mode gets me: "Yep, that’s even better: you get to be lazy and principled at the same time. My favorite combination."

Approaching your LLM with a problem to solve instead of a solution to implement files in the face of modern AI coding best practices. Instead of giving your coding agent a tight spec, starting broad and carving it down to specifics works better with creative LLMs like G3 and Sonnet.

I'm please to announce the release of Sketchy, a hand-drawn, xkcd-inspired design language for Flutter on mobile, desktop, and web: pub.dev/packages/ske... It's my first design system, so your feedback is very welcome!

Google has become the Apple of AI with its expensive but end to end high quality hardware to software to data solution. Nvidia is the come one, come all hardware+framework (CUDA) platform; the Android of AI, if you will. Apple is the Apple of computers and phones but not AI.

As an engineer, I ask every customer coming to me with a solution to step back and talk about the problem first. Likewise, I’m finding that I get better AI results starting with the problem. And by “better,” I mean sometimes crazy better ideas I’d have never come up with on my own.

I'm listening to The Last Economy inside Eleven Reader and loving both. #recommended

The only problem I'm having with Gemini 3 is that I want more! The usage is so high that I'm getting outages. G3 pushes me to be more ambitious, ruining me for other models. It's creative like Claude, competent like Codex but faster than hell. Gemini 3 completes me.

Gemini 3 may be more expensive if you're measuring cost/token. But if you're measuring cost/task and G3 can do things in only a few turns that take the others 2-5x (which is what I'm seeing with my coding tasks), then Gemini 3 is faster, higher quality AND cheaper.

OK, after going around and around with gpt-5.1-codex for an hour, gemini 3 one-shotted it. I was saving my limited access for special occasions, but now g3 is all I wanna use.

I've never built a design system before, but I like how this one is coming along for Flutter.

One of my fun programs ever was figuring out the various parts of a typeface, which I could only do by writing the WinForms program (it was another age : ).

After losing Mint and Personal Capital, and now Empower is moving my cheese again, I needed a stable place to track my accounts. Then I found @SimpleFin (www.simplefin.org). And since I wanted to build my own apps, I built simplefin_dart: pub.dev/packages/sim... Enjoy.

Has anyone tried Zed? I resisted it but am really liking it, especially how they do AI coding agents via the Agent Client Protocol. If it supported Flutter the way VSCode did, I think I'd switch.