Kyle Baxter
@kbaxter.bsky.social
Cause it feels like I've been, I've been here before. The lite brite is now black and white. Exploring LLMs and HCI
http://TightWind.net/
http://TightWind.net/
Agree on it not actually being possible. And I think it’s a responsibility for designers, engineers, data scientists, to have a vision for what they believe is a desirable vision and to build it (myself included). The only way to have a positive future is to build it.
December 11, 2025 at 1:12 AM
Agree on it not actually being possible. And I think it’s a responsibility for designers, engineers, data scientists, to have a vision for what they believe is a desirable vision and to build it (myself included). The only way to have a positive future is to build it.
I’ve moved more or less all of the work work I used to do with o3 and then GPT-5 to Claude with Sonnet 4.5 and Opus. The models are really good and interweaved reasoning is a big deal for me. Also the iOS app’s integration with Calendar and Reminders is useful and underrated
December 11, 2025 at 12:50 AM
I’ve moved more or less all of the work work I used to do with o3 and then GPT-5 to Claude with Sonnet 4.5 and Opus. The models are really good and interweaved reasoning is a big deal for me. Also the iOS app’s integration with Calendar and Reminders is useful and underrated
“The inexorable endpoint is AI automates literally everything” is less a deep, positive analysis of reality than it is a prescription for how AI should be conceptualized and used. But it can be an endpoint if everyone decides, cynically or not, that that’s what it is.
December 11, 2025 at 12:48 AM
“The inexorable endpoint is AI automates literally everything” is less a deep, positive analysis of reality than it is a prescription for how AI should be conceptualized and used. But it can be an endpoint if everyone decides, cynically or not, that that’s what it is.
It’s also a philosophical, what-is-the-good life, question, manifested in design (up and down the stack, from the model to how we interact with it). If no one makes a case (by building) for how AI augments and amplifies what makes us human (in a Licklider sense), that’s a failure.
December 11, 2025 at 12:48 AM
It’s also a philosophical, what-is-the-good life, question, manifested in design (up and down the stack, from the model to how we interact with it). If no one makes a case (by building) for how AI augments and amplifies what makes us human (in a Licklider sense), that’s a failure.
This is great. Endorsed. Resonates with the opportunity atproto presents, too.
December 5, 2025 at 8:40 PM
This is great. Endorsed. Resonates with the opportunity atproto presents, too.
This post seems to run somewhat counter to Anthropic's message that Claude Code (or a harness like it) together with skills is a path toward a generally-applicable agent for many tasks (something I believe is correct). I don't know if I hav a thought behind that, but I find that interesting.
December 3, 2025 at 1:38 AM
This post seems to run somewhat counter to Anthropic's message that Claude Code (or a harness like it) together with skills is a path toward a generally-applicable agent for many tasks (something I believe is correct). I don't know if I hav a thought behind that, but I find that interesting.
Is plan + incremental work completing with "commitable" units + version control for work product generalizable? The post identifies this as an area of exploration. It feels like this can be stated in a more abstract manner that is widely applicable.
December 3, 2025 at 1:35 AM
Is plan + incremental work completing with "commitable" units + version control for work product generalizable? The post identifies this as an area of exploration. It feels like this can be stated in a more abstract manner that is widely applicable.
Some thoughts.
1. I'm not sure this requires a separate "initialization agent," vs instructions for how to plan, and instructions for how to perform development.
2. This feels like a universal problem for a harness/orchestrator working on a task of meaningful complexity for any domain.
1. I'm not sure this requires a separate "initialization agent," vs instructions for how to plan, and instructions for how to perform development.
2. This feels like a universal problem for a harness/orchestrator working on a task of meaningful complexity for any domain.
December 3, 2025 at 1:35 AM
Some thoughts.
1. I'm not sure this requires a separate "initialization agent," vs instructions for how to plan, and instructions for how to perform development.
2. This feels like a universal problem for a harness/orchestrator working on a task of meaningful complexity for any domain.
1. I'm not sure this requires a separate "initialization agent," vs instructions for how to plan, and instructions for how to perform development.
2. This feels like a universal problem for a harness/orchestrator working on a task of meaningful complexity for any domain.
To solve this, they encouraged the agent to do incremental work, commit it, and update the requirement status. This leaves the project in a clean state where a model with a fresh window can pick up a new requirement and the codebase and continue on.
December 3, 2025 at 1:35 AM
To solve this, they encouraged the agent to do incremental work, commit it, and update the requirement status. This leaves the project in a clean state where a model with a fresh window can pick up a new requirement and the codebase and continue on.
The problem they saw is the model wanted to one-shot tasks, leading to poor and incomplete work. They created a first step of drafting a detailed set of requirements to work against; this led to window overflow which compaction didn't solve.
December 3, 2025 at 1:35 AM
The problem they saw is the model wanted to one-shot tasks, leading to poor and incomplete work. They created a first step of drafting a detailed set of requirements to work against; this led to window overflow which compaction didn't solve.
What’s especially unfortunate is we need to keep pulling more of the industry here, but there just isn’t a critical mass (or even near one) that would justify most folks only being here. Contributes to Twitter’s perpetual motion machine.
November 26, 2025 at 8:10 AM
What’s especially unfortunate is we need to keep pulling more of the industry here, but there just isn’t a critical mass (or even near one) that would justify most folks only being here. Contributes to Twitter’s perpetual motion machine.
Thanks! This is the same realization I’ve had over the last 1-2 months too, starting with realizing that agents (with the right combo of prompt and tools, like memory and filesystem) are generalizable (i.e. “deep agents,” but bad term). And then agent skills was like a bolt of lightning
November 22, 2025 at 11:23 PM
Thanks! This is the same realization I’ve had over the last 1-2 months too, starting with realizing that agents (with the right combo of prompt and tools, like memory and filesystem) are generalizable (i.e. “deep agents,” but bad term). And then agent skills was like a bolt of lightning
Those spirits can be given identities which are persistent. Under the hood, there is no “agent,” just an orchestrator, but you can operate with the fiction of them being a tangible persistent thing
November 22, 2025 at 10:59 PM
Those spirits can be given identities which are persistent. Under the hood, there is no “agent,” just an orchestrator, but you can operate with the fiction of them being a tangible persistent thing
(Thinking out loud), where I’m ending up is an agent ecosystem (say, within an enterprise) benefits from an “orchestrator” primitive, whereas with 1:1 spirit:body pairing wasn’t req’d. Orchestrator executes a spirit, whatever complexity it has (ranging in analogy from single agent to multi-agent).
November 22, 2025 at 10:59 PM
(Thinking out loud), where I’m ending up is an agent ecosystem (say, within an enterprise) benefits from an “orchestrator” primitive, whereas with 1:1 spirit:body pairing wasn’t req’d. Orchestrator executes a spirit, whatever complexity it has (ranging in analogy from single agent to multi-agent).
In that sense, the “body” is infinitely copyable, and is more an actuator in an env for the spirit. But we still need the spirit+body to reason about what are the actors within an environment (I think). But it feels weird given that separation of spirit and body to recreate “agent” for that purpose
November 22, 2025 at 10:39 PM
In that sense, the “body” is infinitely copyable, and is more an actuator in an env for the spirit. But we still need the spirit+body to reason about what are the actors within an environment (I think). But it feels weird given that separation of spirit and body to recreate “agent” for that purpose
imo this distinction is relatively new and really important. “Agent” before was usually a singular combo of what you refer to as the spirit and body; now the spirit is separately definable and loadable. I am still struggling tbh with what is the primary unit to reason about, spirit or body
November 22, 2025 at 10:37 PM
imo this distinction is relatively new and really important. “Agent” before was usually a singular combo of what you refer to as the spirit and body; now the spirit is separately definable and loadable. I am still struggling tbh with what is the primary unit to reason about, spirit or body
They hit on this in a post a few weeks ago: www.anthropic.com/engineering/... Skills really does feel like a durable phase change on top of model+tools
Code execution with MCP: building more efficient AI agents
Learn how code execution with the Model Context Protocol enables agents to handle more tools while using fewer tokens, reducing context overhead by up to 98.7%.
www.anthropic.com
November 22, 2025 at 3:06 PM
They hit on this in a post a few weeks ago: www.anthropic.com/engineering/... Skills really does feel like a durable phase change on top of model+tools
Yeah, that’s what’s captured my thoughts for the last few weeks. I’ve been experimenting with having Claude write skills for various things using primary source documents and it’s done a quite good job. A model doing that to formalize a task path seems very doable, and yeah, it is continual learning
November 22, 2025 at 2:59 PM
Yeah, that’s what’s captured my thoughts for the last few weeks. I’ve been experimenting with having Claude write skills for various things using primary source documents and it’s done a quite good job. A model doing that to formalize a task path seems very doable, and yeah, it is continual learning
Any new takeaways from this talk re skills? I’ve been pretty fascinated with their concept of “tool” being a script the model retrieves (rather than a function call) and runs in a sandbox, with the model able to also formalize paths it takes to solve tasks as skills in its sandbox
November 22, 2025 at 5:10 AM
Any new takeaways from this talk re skills? I’ve been pretty fascinated with their concept of “tool” being a script the model retrieves (rather than a function call) and runs in a sandbox, with the model able to also formalize paths it takes to solve tasks as skills in its sandbox