Chris Sells
@csells.bsky.social
Chris Sells is an independent consultant focused on applied AI and developer tools and ecosystems. He enjoys long walks on the beach and various technologies.
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.
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.
SimpleFIN
SimpleFIN is a protocol for simple financial aggregation.
www.simplefin.org
November 3, 2025 at 4:32 AM
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.
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.
October 31, 2025 at 8:01 PM
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.
dartantic_ai v1.3.0 release @ pub.dev/packages/dar...:
- Anthropic thinking support exposed like the OpenAI Responses provider, so you can write your code to look for thinking output regardless of the provider
- Ollama typed output support
- Mistral usage tracking
- Anthropic thinking support exposed like the OpenAI Responses provider, so you can write your code to look for thinking output regardless of the provider
- Ollama typed output support
- Mistral usage tracking
dartantic_ai | Dart package
Dartantic is an agentic framework designed to make building client and server-side apps in Dart with generative AI easier and more fun!
pub.dev
October 30, 2025 at 2:01 PM
dartantic_ai v1.3.0 release @ pub.dev/packages/dar...:
- Anthropic thinking support exposed like the OpenAI Responses provider, so you can write your code to look for thinking output regardless of the provider
- Ollama typed output support
- Mistral usage tracking
- Anthropic thinking support exposed like the OpenAI Responses provider, so you can write your code to look for thinking output regardless of the provider
- Ollama typed output support
- Mistral usage tracking
I shipped dartantic_ai 1.2.0: pub.dev/packages/dar...
- Migrated Google provider to the new Google-supported
package, fixing some low-level response issues.
- Added Google "double agent" orchestrator to support typed output with tools simultaneously.
- Added llms.txt for LLM readers.
- And more!
- Migrated Google provider to the new Google-supported
package, fixing some low-level response issues.
- Added Google "double agent" orchestrator to support typed output with tools simultaneously.
- Added llms.txt for LLM readers.
- And more!
dartantic_ai | Dart package
Dartantic is an agentic framework designed to make building client and server-side apps in Dart with generative AI easier and more fun!
pub.dev
October 29, 2025 at 3:36 PM
I shipped dartantic_ai 1.2.0: pub.dev/packages/dar...
- Migrated Google provider to the new Google-supported
package, fixing some low-level response issues.
- Added Google "double agent" orchestrator to support typed output with tools simultaneously.
- Added llms.txt for LLM readers.
- And more!
- Migrated Google provider to the new Google-supported
package, fixing some low-level response issues.
- Added Google "double agent" orchestrator to support typed output with tools simultaneously.
- Added llms.txt for LLM readers.
- And more!
Been using AI coding agents for a while? Then watch this video to have Lada Kesseler turn your niggling suspicions into "I knew it!"s: www.youtube.com/watch?v=_LSK...
Lada has pulled out the Gamma-style patterns out of coding agents. #recommended
Lada has pulled out the Gamma-style patterns out of coding agents. #recommended
AI Talks - Lada Kesseler: Augmented Coding: Mapping the Uncharted Territory
YouTube video by Calgary Software Crafters
www.youtube.com
October 28, 2025 at 1:46 AM
Been using AI coding agents for a while? Then watch this video to have Lada Kesseler turn your niggling suspicions into "I knew it!"s: www.youtube.com/watch?v=_LSK...
Lada has pulled out the Gamma-style patterns out of coding agents. #recommended
Lada has pulled out the Gamma-style patterns out of coding agents. #recommended
Interested in how dartantic does all of the cool things it does? Check out the wiki: github.com/csells/darta...
Home
Dartantic is an agentic framework designed to make building client and server-side apps in Dart with generative AI easier and more fun! - csells/dartantic_ai
github.com
October 14, 2025 at 12:49 AM
Interested in how dartantic does all of the cool things it does? Check out the wiki: github.com/csells/darta...
Dartantic v1.1.0 is out and this is a big release!
New OpenAI Responses provider, including session persistence (aka prompt caching), intrinsic server-side tools, and thinking metadata streams.
Thanks to @jezell.bsky.social for his openai_core package!
Check out dartantic.ai for the deets.
New OpenAI Responses provider, including session persistence (aka prompt caching), intrinsic server-side tools, and thinking metadata streams.
Thanks to @jezell.bsky.social for his openai_core package!
Check out dartantic.ai for the deets.
Redirecting to docs.dartantic.ai
dartantic.ai
October 10, 2025 at 12:41 AM
Dartantic v1.1.0 is out and this is a big release!
New OpenAI Responses provider, including session persistence (aka prompt caching), intrinsic server-side tools, and thinking metadata streams.
Thanks to @jezell.bsky.social for his openai_core package!
Check out dartantic.ai for the deets.
New OpenAI Responses provider, including session persistence (aka prompt caching), intrinsic server-side tools, and thinking metadata streams.
Thanks to @jezell.bsky.social for his openai_core package!
Check out dartantic.ai for the deets.
Interested in the core issues in building a real-world AI coding agent? Start with @pvncher's amazing blog post: repoprompt.com/blog/buildin...
Building Repo Bench – Repo Prompt Blog
How I built a benchmark to measure what truly matters for AI coding assistants - file editing precision, instruction following, and handling complex context.
repoprompt.com
October 8, 2025 at 11:38 PM
Interested in the core issues in building a real-world AI coding agent? Start with @pvncher's amazing blog post: repoprompt.com/blog/buildin...
I'm busy working on a new dartantic release and as part of that, Sonnet 4.5 runs examples like this:
`DARTANTIC_LOG_LEVEL=FINE dart run server_side_vector_search.dart`
This was surprising, since dartantic had no environment check for a logging level.
But now it does. : )
`DARTANTIC_LOG_LEVEL=FINE dart run server_side_vector_search.dart`
This was surprising, since dartantic had no environment check for a logging level.
But now it does. : )
October 5, 2025 at 8:12 PM
I'm busy working on a new dartantic release and as part of that, Sonnet 4.5 runs examples like this:
`DARTANTIC_LOG_LEVEL=FINE dart run server_side_vector_search.dart`
This was surprising, since dartantic had no environment check for a logging level.
But now it does. : )
`DARTANTIC_LOG_LEVEL=FINE dart run server_side_vector_search.dart`
This was surprising, since dartantic had no environment check for a logging level.
But now it does. : )
Been spending a bunch of time with Gemini 2.5 Pro for coding this weekend.
It's about 80% as smart for coding as GPT5 and Opus 4, but when I use it, I can't juggle between multiple agents on different projects:
2.5 Pro is too fast!
Not a bad tradeoff to be honest.
It's about 80% as smart for coding as GPT5 and Opus 4, but when I use it, I can't juggle between multiple agents on different projects:
2.5 Pro is too fast!
Not a bad tradeoff to be honest.
September 21, 2025 at 10:48 PM
Been spending a bunch of time with Gemini 2.5 Pro for coding this weekend.
It's about 80% as smart for coding as GPT5 and Opus 4, but when I use it, I can't juggle between multiple agents on different projects:
2.5 Pro is too fast!
Not a bad tradeoff to be honest.
It's about 80% as smart for coding as GPT5 and Opus 4, but when I use it, I can't juggle between multiple agents on different projects:
2.5 Pro is too fast!
Not a bad tradeoff to be honest.
Holy cow. Somebody thinks that I'm in the top 1% Dart engineers in the world. I do love Dart!
algora.io/profile/csells
algora.io/profile/csells
csells | Algora
Chris Sells is an advocate for developers. He enjoys long walks on the beach and various technologies.
algora.io
September 18, 2025 at 12:49 AM
Holy cow. Somebody thinks that I'm in the top 1% Dart engineers in the world. I do love Dart!
algora.io/profile/csells
algora.io/profile/csells
I used to get up in the morning and deal with my messages.
Now I check the progress of my AI agents and direct them to the next step, while my messages go unmanaged.
AI just feels more creative and fun.
Now I check the progress of my AI agents and direct them to the next step, while my messages go unmanaged.
AI just feels more creative and fun.
September 15, 2025 at 3:34 PM
I used to get up in the morning and deal with my messages.
Now I check the progress of my AI agents and direct them to the next step, while my messages go unmanaged.
AI just feels more creative and fun.
Now I check the progress of my AI agents and direct them to the next step, while my messages go unmanaged.
AI just feels more creative and fun.
TIL not to use AI to code a protocol client and server in the same project at the same time, as it will surely route around the protocol restrictions to get things working.
Instead, build the new server/client against known working implementations first.
This is the way.
Instead, build the new server/client against known working implementations first.
This is the way.
September 9, 2025 at 3:07 PM
TIL not to use AI to code a protocol client and server in the same project at the same time, as it will surely route around the protocol restrictions to get things working.
Instead, build the new server/client against known working implementations first.
This is the way.
Instead, build the new server/client against known working implementations first.
This is the way.
Want to know how to get the best from agentic coding agents? Matt's video represents the best explanation for getting solid results that I've seen (aka this is what I do so it must be right : )
www.youtube.com/watch?v=dBWn...
www.youtube.com/watch?v=dBWn...
Stop Letting AI Wing It -- Take Control of AI Coding
YouTube video by Matt Maher
www.youtube.com
September 7, 2025 at 12:53 AM
Want to know how to get the best from agentic coding agents? Matt's video represents the best explanation for getting solid results that I've seen (aka this is what I do so it must be right : )
www.youtube.com/watch?v=dBWn...
www.youtube.com/watch?v=dBWn...
I think I said to an AI coding agent today: "you're fixing the test to not require the thing we're testing for? why would you betray me like that? do you hate america?"
the reply: "Ha — I promise I don’t hate America. I hate flaky tests."
who says you can't make work fun? : )
the reply: "Ha — I promise I don’t hate America. I hate flaky tests."
who says you can't make work fun? : )
September 5, 2025 at 4:12 AM
I think I said to an AI coding agent today: "you're fixing the test to not require the thing we're testing for? why would you betray me like that? do you hate america?"
the reply: "Ha — I promise I don’t hate America. I hate flaky tests."
who says you can't make work fun? : )
the reply: "Ha — I promise I don’t hate America. I hate flaky tests."
who says you can't make work fun? : )
Cline is continuing to dig into solving real-world problems with heavy duty AI agents:
cline.bot/blog/cline-v...
cline.bot/blog/cline-v...
Cline v3.25: The Coding Agent Built for Hard Problems - Cline Blog
In v3.25, Cline never lets go of the thread.
cline.bot
August 19, 2025 at 9:56 PM
Cline is continuing to dig into solving real-world problems with heavy duty AI agents:
cline.bot/blog/cline-v...
cline.bot/blog/cline-v...
"We aim to emphasize high-value tokens: where every step in a plan is deliberate, relevant, and positioned where the model can use it. By anchoring the agent on what matters most, we make better use of every token — and focus the model’s attention."
cline.bot/blog/focus-a...
cline.bot/blog/focus-a...
Focus: attention isn’t enough - Cline Blog
Bigger context windows don’t stop coding agents from drifting. Cline’s new Focus Chain keeps every step anchored to the goal — so even the longest tasks finish exactly where they’re meant to.
cline.bot
August 15, 2025 at 4:22 PM
"We aim to emphasize high-value tokens: where every step in a plan is deliberate, relevant, and positioned where the model can use it. By anchoring the agent on what matters most, we make better use of every token — and focus the model’s attention."
cline.bot/blog/focus-a...
cline.bot/blog/focus-a...
Welcome to dartantic_ai 1.0!
- 11 providers (Google, OpenAI, Anthropic, ...)
- Tool calling with typed I/O
- Runs everywhere Dart runs
sellsbrothers.com/welcome-to-d...
- 11 providers (Google, OpenAI, Anthropic, ...)
- Tool calling with typed I/O
- Runs everywhere Dart runs
sellsbrothers.com/welcome-to-d...
Welcome to dartantic_ai 1.0! - Marquee de Sells: Chris's insight outlet
Dartantic is an agentic framework designed to make building client and server-side apps in Dart with generative AI easier and more fun! It works
sellsbrothers.com
July 29, 2025 at 6:12 PM
Welcome to dartantic_ai 1.0!
- 11 providers (Google, OpenAI, Anthropic, ...)
- Tool calling with typed I/O
- Runs everywhere Dart runs
sellsbrothers.com/welcome-to-d...
- 11 providers (Google, OpenAI, Anthropic, ...)
- Tool calling with typed I/O
- Runs everywhere Dart runs
sellsbrothers.com/welcome-to-d...
I've had some early experience with the Dart MCP Server and it's got a ton of potential for attaching AI agents to your running Flutter apps: medium.com/flutter/supe...
Supercharge Your Dart & Flutter Development Experience with the Dart MCP Server
By Ander Dobo — Product Manager, and Kenzie Davisson — Engineer
medium.com
July 24, 2025 at 1:55 AM
I've had some early experience with the Dart MCP Server and it's got a ton of potential for attaching AI agents to your running Flutter apps: medium.com/flutter/supe...
I have discovered that I am as picky as any artist with their work.
Dartantic 1.0 was functionally complete and passing >1000 tests across 11 chat model providers and 4 embeddings model providers three days ago.
But I couldn't ship it because it wasn't *right*.
Dartantic 1.0 was functionally complete and passing >1000 tests across 11 chat model providers and 4 embeddings model providers three days ago.
But I couldn't ship it because it wasn't *right*.
July 22, 2025 at 11:17 PM
I have discovered that I am as picky as any artist with their work.
Dartantic 1.0 was functionally complete and passing >1000 tests across 11 chat model providers and 4 embeddings model providers three days ago.
But I couldn't ship it because it wasn't *right*.
Dartantic 1.0 was functionally complete and passing >1000 tests across 11 chat model providers and 4 embeddings model providers three days ago.
But I couldn't ship it because it wasn't *right*.
The 5 Stages of AI Grief: sellsbrothers.com/the-5-stages...
The 5 Stages of AI Grief - Marquee de Sells: Chris's insight outlet
“The future is already here — it’s just not very evenly distributed.” –W. Gibson As a consultant and speaker, I talk to a lot of software
sellsbrothers.com
July 22, 2025 at 2:18 PM
The 5 Stages of AI Grief: sellsbrothers.com/the-5-stages...
"If something must happen, please don't use a stochastic process to do it. Eventually, due to the nature of sampling from a probability distribution, it will fail to do the thing that must be done." @domesticmouse.com
July 22, 2025 at 12:27 PM
"If something must happen, please don't use a stochastic process to do it. Eventually, due to the nature of sampling from a probability distribution, it will fail to do the thing that must be done." @domesticmouse.com
I've been talking to the Cactus folks about getting their local LLMs for mobile into dartantic. In the meantime, don't let that stop you from bringing them into your mobile Flutter apps: medium.com/@shemet0roma...
How to run private, on-device AI in your Flutter App using Cactus
Deploy private, offline, on-device AI in your Flutter app in a few lines of code.
medium.com
July 18, 2025 at 2:51 AM
I've been talking to the Cactus folks about getting their local LLMs for mobile into dartantic. In the meantime, don't let that stop you from bringing them into your mobile Flutter apps: medium.com/@shemet0roma...
I'm happy to announce flutter_ai_toolkit v0.9.2. This release slays the remaining bugs on the road to v1.0 plus:
- multi-turn tool call support
- support for uploading attachments to the cloud
- overriding the speech-to-text conversion
pub.dev/packages/flu...
- multi-turn tool call support
- support for uploading attachments to the cloud
- overriding the speech-to-text conversion
pub.dev/packages/flu...
flutter_ai_toolkit changelog | Flutter package
A set of AI chat-related widgets for Flutter apps targeting mobile, desktop, and web.
pub.dev
July 17, 2025 at 10:25 PM
I'm happy to announce flutter_ai_toolkit v0.9.2. This release slays the remaining bugs on the road to v1.0 plus:
- multi-turn tool call support
- support for uploading attachments to the cloud
- overriding the speech-to-text conversion
pub.dev/packages/flu...
- multi-turn tool call support
- support for uploading attachments to the cloud
- overriding the speech-to-text conversion
pub.dev/packages/flu...