verdverm
@verdverm.com
dev & entrepreneur interested in atproto, cuelang, machine learning, developer experience, combating misinformation
working on https://blebbit.app | @blebbit.app | #blebbit
personal: https://verdverm.com | https://github.com/verdverm
working on https://blebbit.app | @blebbit.app | #blebbit
personal: https://verdverm.com | https://github.com/verdverm
ok, this is freaking sweet
Every chat session gets an isolated filesystem in Dagger. With VS Code virtual filesystems, the difference is transparent to you.
Co-edit files with chat, both of you see the changes so you can
1. generate some code
2. manually fix an error
3. give back to agent
Every chat session gets an isolated filesystem in Dagger. With VS Code virtual filesystems, the difference is transparent to you.
Co-edit files with chat, both of you see the changes so you can
1. generate some code
2. manually fix an error
3. give back to agent
November 28, 2025 at 6:13 AM
ok, this is freaking sweet
Every chat session gets an isolated filesystem in Dagger. With VS Code virtual filesystems, the difference is transparent to you.
Co-edit files with chat, both of you see the changes so you can
1. generate some code
2. manually fix an error
3. give back to agent
Every chat session gets an isolated filesystem in Dagger. With VS Code virtual filesystems, the difference is transparent to you.
Co-edit files with chat, both of you see the changes so you can
1. generate some code
2. manually fix an error
3. give back to agent
My agents are far enough along that they are starting to do real work
There must be a free quota per day, because my billing has barely moved since my $5 mistake
There must be a free quota per day, because my billing has barely moved since my $5 mistake
November 27, 2025 at 7:16 AM
My agents are far enough along that they are starting to do real work
There must be a free quota per day, because my billing has barely moved since my $5 mistake
There must be a free quota per day, because my billing has barely moved since my $5 mistake
I haven't tried it yet, want to try something almost the opposite direction first
I saw in the supposed claude prompts they give problem solving approaches and examples of query->plan->calls->result kinds of things
I'm building a bunch of what I call prompt fragments, probably like skills and such
I saw in the supposed claude prompts they give problem solving approaches and examples of query->plan->calls->result kinds of things
I'm building a bunch of what I call prompt fragments, probably like skills and such
November 27, 2025 at 4:16 AM
I haven't tried it yet, want to try something almost the opposite direction first
I saw in the supposed claude prompts they give problem solving approaches and examples of query->plan->calls->result kinds of things
I'm building a bunch of what I call prompt fragments, probably like skills and such
I saw in the supposed claude prompts they give problem solving approaches and examples of query->plan->calls->result kinds of things
I'm building a bunch of what I call prompt fragments, probably like skills and such
Added the #veg template system to #veg #agentic.
Define, compose, and condition instructions based on model, agent config, or any state value
...per chat turn, with only markdown fragments + text/template
...with modules, registries, and sharing out of the (pandora's) box
also 4 agents & tools
Define, compose, and condition instructions based on model, agent config, or any state value
...per chat turn, with only markdown fragments + text/template
...with modules, registries, and sharing out of the (pandora's) box
also 4 agents & tools
November 27, 2025 at 3:02 AM
I present #agentic session time travel built on @dagger.io
coming soon, support for from any container, exec any command, push / pull / clone / branch / fork
agents, tools, and instructions will be joining the ride
coming soon, support for from any container, exec any command, push / pull / clone / branch / fork
agents, tools, and instructions will be joining the ride
November 26, 2025 at 7:00 PM
I present #agentic session time travel built on @dagger.io
coming soon, support for from any container, exec any command, push / pull / clone / branch / fork
agents, tools, and instructions will be joining the ride
coming soon, support for from any container, exec any command, push / pull / clone / branch / fork
agents, tools, and instructions will be joining the ride
my hack of an #agentic coder made its first real code contribution to itself, fixing a slowdown when chat sessions get long (100+ events)
the long chat was testing a fix for the 2.5M token ($5) mistake I made the other day, blowing up the context
the long chat was testing a fix for the 2.5M token ($5) mistake I made the other day, blowing up the context
November 25, 2025 at 12:36 PM
my hack of an #agentic coder made its first real code contribution to itself, fixing a slowdown when chat sessions get long (100+ events)
the long chat was testing a fix for the 2.5M token ($5) mistake I made the other day, blowing up the context
the long chat was testing a fix for the 2.5M token ($5) mistake I made the other day, blowing up the context
It's time to bring in some @cuelang.org for my #agent
Initially this is for me, but...
what if I give an agent the ability to dynamically create other agents by calling this helper function wrapped in an agent with a touch of CUE, JSON, or Yaml?
Initially this is for me, but...
what if I give an agent the ability to dynamically create other agents by calling this helper function wrapped in an agent with a touch of CUE, JSON, or Yaml?
November 23, 2025 at 4:12 AM
It's time to bring in some @cuelang.org for my #agent
Initially this is for me, but...
what if I give an agent the ability to dynamically create other agents by calling this helper function wrapped in an agent with a touch of CUE, JSON, or Yaml?
Initially this is for me, but...
what if I give an agent the ability to dynamically create other agents by calling this helper function wrapped in an agent with a touch of CUE, JSON, or Yaml?
GitHub has an off-by-one, spam notification, that I cannot get rid of...
it's things like this that make me hate everything about the crypto bros and their "ecosystem" of takers and fakers
it's things like this that make me hate everything about the crypto bros and their "ecosystem" of takers and fakers
November 21, 2025 at 10:36 PM
GitHub has an off-by-one, spam notification, that I cannot get rid of...
it's things like this that make me hate everything about the crypto bros and their "ecosystem" of takers and fakers
it's things like this that make me hate everything about the crypto bros and their "ecosystem" of takers and fakers
The Gemini family, some of the following are single LLM, others are starting my real agentic journey. I've set it up so that it is easy to make and try out lots of variations.
The plan is to move agent definitions to CUE, build out some eval capabilities, and record everything locally
adk-go 👨🍳 💋
The plan is to move agent definitions to CUE, build out some eval capabilities, and record everything locally
adk-go 👨🍳 💋
November 20, 2025 at 8:55 AM
The Gemini family, some of the following are single LLM, others are starting my real agentic journey. I've set it up so that it is easy to make and try out lots of variations.
The plan is to move agent definitions to CUE, build out some eval capabilities, and record everything locally
adk-go 👨🍳 💋
The plan is to move agent definitions to CUE, build out some eval capabilities, and record everything locally
adk-go 👨🍳 💋
This is like 4-5 days work with ADK in VS Code
(mostly vs code, I can turn back to ADK for a bit, gemini 3 is calling)
(mostly vs code, I can turn back to ADK for a bit, gemini 3 is calling)
November 19, 2025 at 4:17 AM
This is like 4-5 days work with ADK in VS Code
(mostly vs code, I can turn back to ADK for a bit, gemini 3 is calling)
(mostly vs code, I can turn back to ADK for a bit, gemini 3 is calling)
Interesting, I added similar features to my custom agent yesterday because I want deeper integration into the planning / assigning worker parts of agentic coding.
Iterative, not one-shot, like we do IRL with other humans
Will have to see what else I can glean for ideation
Iterative, not one-shot, like we do IRL with other humans
Will have to see what else I can glean for ideation
November 18, 2025 at 10:51 PM
Interesting, I added similar features to my custom agent yesterday because I want deeper integration into the planning / assigning worker parts of agentic coding.
Iterative, not one-shot, like we do IRL with other humans
Will have to see what else I can glean for ideation
Iterative, not one-shot, like we do IRL with other humans
Will have to see what else I can glean for ideation
So glad I stayed up until 4am working on my custom agent ui. Turns out it was JIT for Gemini 3
November 18, 2025 at 6:17 PM
So glad I stayed up until 4am working on my custom agent ui. Turns out it was JIT for Gemini 3
VS Code extension coming along, have it hooked up to the agent server over websocket and json payloads so we can do way more than just chat messages
Also pleased to learn I can use react, tanstack, and tailwind
Soon it will be building itself and private data for @atproto.com
Also pleased to learn I can use react, tanstack, and tailwind
Soon it will be building itself and private data for @atproto.com
November 16, 2025 at 8:01 AM
VS Code extension coming along, have it hooked up to the agent server over websocket and json payloads so we can do way more than just chat messages
Also pleased to learn I can use react, tanstack, and tailwind
Soon it will be building itself and private data for @atproto.com
Also pleased to learn I can use react, tanstack, and tailwind
Soon it will be building itself and private data for @atproto.com
Started a new VS Code extension today for a custom agent. One of the features that pushed me to do this is having the agent be aware of and able to interact with any terminal I have open. Current options run in or start their own terminal, and cannot jump into my current work.
November 15, 2025 at 10:13 PM
Started a new VS Code extension today for a custom agent. One of the features that pushed me to do this is having the agent be aware of and able to interact with any terminal I have open. Current options run in or start their own terminal, and cannot jump into my current work.
chars & graphemes are distinct things (ascii vs unicode aiui)
November 14, 2025 at 2:42 AM
chars & graphemes are distinct things (ascii vs unicode aiui)
Maybe the meme needs a refresh from the ubiquity of the attention economy, based on something like this
November 14, 2025 at 1:07 AM
Maybe the meme needs a refresh from the ubiquity of the attention economy, based on something like this
@kelseyhightower.com this your doing?
November 14, 2025 at 12:03 AM
@kelseyhightower.com this your doing?
More generally this applies to unlocking query parameters for a number of interesting and unforeseen possibilities
November 13, 2025 at 10:42 PM
More generally this applies to unlocking query parameters for a number of interesting and unforeseen possibilities
South Park has been on point this year
November 12, 2025 at 7:00 PM
South Park has been on point this year
Democrats have nearly 2x the number of members over 70 as compared to Republicans
www.nytimes.com/interactive/...
@schumer.senate.gov YOU FAILED US AND NEED TO GO!
www.nytimes.com/interactive/...
@schumer.senate.gov YOU FAILED US AND NEED TO GO!
November 11, 2025 at 12:04 AM
Democrats have nearly 2x the number of members over 70 as compared to Republicans
www.nytimes.com/interactive/...
@schumer.senate.gov YOU FAILED US AND NEED TO GO!
www.nytimes.com/interactive/...
@schumer.senate.gov YOU FAILED US AND NEED TO GO!