Ben Davis's banner
Ben Davis's profile picture

Ben Davis

@davis7 • 17,488 subscribers

yells at @theo both on and off camera (co-host of the nerd snipe podcast)

Shorts

The new btca experience, now just a skill b/c it's way better: Turns out the best way to let coding agents search github for context is to let them search github repos for context

The new btca experience, now just a skill b/c it's way better: Turns out the best way to let coding agents search github for context is to let them search github repos for context

87,820 views

Michael Arnaldi suggested cloning the effect repo as a git subtree, giving it to Claude, then using that as docs It sounds absurd, but it's actually kind amazing lol Setup a custom OpenCode agent that has all the context on where the files are then it kinda just works I feel like this could definitely be refined but idk if I even care enough. Stuff like cursor/opencode/claude code is already so good at searching codebases that I don't really care to do anything more

Michael Arnaldi suggested cloning the effect repo as a git subtree, giving it to Claude, then using that as docs It sounds absurd, but it's actually kind amazing lol Setup a custom OpenCode agent that has all the context on where the files are then it kinda just works I feel like this could definitely be refined but idk if I even care enough. Stuff like cursor/opencode/claude code is already so good at searching codebases that I don't really care to do anything more

112,524 views

Updated the subagents system in my custom Pi extension to include: - Pi subagents - Codex subagents - Claude Code subagents The Pi thread can call and spin up any of them, and instead of just using the "codex exec" or whatever, it'll do a tool call that wraps it up in a first class way. Have a decent UI for them already added in as well. Pi being the main thread then having Codex/Claude Code as tools when needed feels so good. A flow I'm already using a ton is: "Implement feature ___, have a cc fable subagent research and plan out the api design + feature, then have a codex subagent implement it, then have a cc fable subagent do a review/fixes focused on simplicity and correctness" I also have a dynamic workflows extension setup and working, but it's not nearly as polished yet. Probably gonna make the subagent system more robust, then built it around that so u could have a workflow that flows between codex/cc/pi...

Updated the subagents system in my custom Pi extension to include: - Pi subagents - Codex subagents - Claude Code subagents The Pi thread can call and spin up any of them, and instead of just using the "codex exec" or whatever, it'll do a tool call that wraps it up in a first class way. Have a decent UI for them already added in as well. Pi being the main thread then having Codex/Claude Code as tools when needed feels so good. A flow I'm already using a ton is: "Implement feature ___, have a cc fable subagent research and plan out the api design + feature, then have a codex subagent implement it, then have a cc fable subagent do a review/fixes focused on simplicity and correctness" I also have a dynamic workflows extension setup and working, but it's not nearly as polished yet. Probably gonna make the subagent system more robust, then built it around that so u could have a workflow that flows between codex/cc/pi...

24,827 views

Can finally play doom in Claude Code

Can finally play doom in Claude Code

36,817 views

Videos

LIVE
1.2k

Anya Rossi

sweetdream.ai

SweetDream.aiSponsored•Livecam
Streaming Now

Watch Anya Live

Anya is streaming live right now! Join her private show and enjoy exclusive content.

HD live stream
Exclusive private shows
1.2k viewers online

Free preview available • Premium content

This is very late, but I'm finally done with my 5.5 vid - use low reasoning - the name sucks - it's fast - best code I've ever seen a model write came from this model - openai's new pre-training is amazing - price looks worse than it is - over sensitive to every little thing in it's context window - feels wildly different compared to 5.4 - turn reasoning off. try it. turn the reasoning off. do it.
28:50
davis7's profile picture

This is very late, but I'm finally done with my 5.5 vid - use low reasoning - the name sucks - it's fast - best code I've ever seen a model write came from this model - openai's new pre-training is amazing - price looks worse than it is - over sensitive to every little thing in it's context window - feels wildly different compared to 5.4 - turn reasoning off. try it. turn the reasoning off. do it.

Ben Davis

283,972 views • 4 months ago

My Opus 5 video is late, but that ended up being good because I would have made a very different video on release day I really like the idea of this model, but actually using it kinda sucks...
16:34
davis7's profile picture

My Opus 5 video is late, but that ended up being good because I would have made a very different video on release day I really like the idea of this model, but actually using it kinda sucks...

Ben Davis

57,560 views • 1 month ago

Made a video on my custom Pi extensions: - Subagents (Pi, Codex, Claude Code) - Workflows - Question ask tool - Background Terminals - Custom UI and more Combining all the harnesses together has been one of my favorite ways to work, Pi feels like the best way to do it...
30:34
davis7's profile picture

Made a video on my custom Pi extensions: - Subagents (Pi, Codex, Claude Code) - Workflows - Question ask tool - Background Terminals - Custom UI and more Combining all the harnesses together has been one of my favorite ways to work, Pi feels like the best way to do it...

Ben Davis

41,480 views • 1 month ago

OpenAI and Anthropic finally have real competition. The SpaceXAi comeback has been wild to watch, really hope it continues and we get even more competition in the space
26:04
davis7's profile picture

OpenAI and Anthropic finally have real competition. The SpaceXAi comeback has been wild to watch, really hope it continues and we get even more competition in the space

Ben Davis

14,012 views • 13 days ago

Put together a vid on how/why I'm using Hermes agent My networking setup is very overkill for most people, I mostly did it because I'm paranoid (and tbh it was fun) But it's really useful, highly recommend setting one up it's a great piece of software
24:40
davis7's profile picture

Put together a vid on how/why I'm using Hermes agent My networking setup is very overkill for most people, I mostly did it because I'm paranoid (and tbh it was fun) But it's really useful, highly recommend setting one up it's a great piece of software

Ben Davis

56,002 views • 2 months ago

Made a video on how I decide between models right now. Still trying to figure out the exact way to describe these things, especially as the agents get more capable, but at least for now this is how I'm thinking about things (80% GPT-5.6 Sol, 15% Fable, 5% everything else)
21:25
davis7's profile picture

Made a video on how I decide between models right now. Still trying to figure out the exact way to describe these things, especially as the agents get more capable, but at least for now this is how I'm thinking about things (80% GPT-5.6 Sol, 15% Fable, 5% everything else)

Ben Davis

11,684 views • 19 days ago

A video I never thought I would make: defending GStack I honestly wrote it off as a meme, but the concept is actually really really cool Idk exactly what this turns into long term, but the idea of "markdown file == program" has so much unexplored potential
23:56
davis7's profile picture

A video I never thought I would make: defending GStack I honestly wrote it off as a meme, but the concept is actually really really cool Idk exactly what this turns into long term, but the idea of "markdown file == program" has so much unexplored potential

Ben Davis

46,937 views • 3 months ago

Haiku 4.5 has finally been dethroned by GPT 5.4 mini: - faster - cheaper - smarter - better at tool calling Gemini 3.0 flash still has a place in “turn this video/audio/blob of text into json”, but for any small model task that requires an agent loop (sub agents, search, quick fixes, browser use, etc.) GPT-5.4 mini is the new king
11:04
davis7's profile picture

Haiku 4.5 has finally been dethroned by GPT 5.4 mini: - faster - cheaper - smarter - better at tool calling Gemini 3.0 flash still has a place in “turn this video/audio/blob of text into json”, but for any small model task that requires an agent loop (sub agents, search, quick fixes, browser use, etc.) GPT-5.4 mini is the new king

Ben Davis

67,624 views • 5 months ago

it went about as expected
0:50
davis7's profile picture

it went about as expected

Ben Davis

40,857 views • 4 months ago

Been testing 5.4 for about a week now and it's really good. It's the first model that kinda "just works". A hybrid of 5.3's precision and opus' speed/initiative I've done ~300 million tokens with it and it's crushed almost everything I've given it. Except for UI. It's hilariously bad at UI stuff.
28:28
davis7's profile picture

Been testing 5.4 for about a week now and it's really good. It's the first model that kinda "just works". A hybrid of 5.3's precision and opus' speed/initiative I've done ~300 million tokens with it and it's crushed almost everything I've given it. Except for UI. It's hilariously bad at UI stuff.

Ben Davis

54,357 views • 6 months ago

I was wrong about GPT-5.5 Kinda I still stand by everything I've said about low reasoning, it's incredible and I still use it all the time esp with Pi But I'm also doing a ton with xHigh reasoning, and especially computer use in the codex desktop app
17:33
davis7's profile picture

I was wrong about GPT-5.5 Kinda I still stand by everything I've said about low reasoning, it's incredible and I still use it all the time esp with Pi But I'm also doing a ton with xHigh reasoning, and especially computer use in the codex desktop app

Ben Davis

30,205 views • 3 months ago

I've spent a lot of time customizing my Pi setup It's quickly become my favorite coding agent, u just need to make it yours Made a vid on how (TL;DR, just tell it to change itself and it will):
21:26
davis7's profile picture

I've spent a lot of time customizing my Pi setup It's quickly become my favorite coding agent, u just need to make it yours Made a vid on how (TL;DR, just tell it to change itself and it will):

Ben Davis

31,263 views • 3 months ago

I'm tired and need to go to bed, it's 4AM, but it's live: - setup as many git repos as "resources" as you want - "@" them in the chat UI and haiku 4.5 will use them to answer your question - $8/mo. for the hosted version, cli is BYOK - open source
0:23
davis7's profile picture

I'm tired and need to go to bed, it's 4AM, but it's live: - setup as many git repos as "resources" as you want - "@" them in the chat UI and haiku 4.5 will use them to answer your question - $8/mo. for the hosted version, cli is BYOK - open source

Ben Davis

54,300 views • 7 months ago

made a video b/c Theo was wrong
23:01
davis7's profile picture

made a video b/c Theo was wrong

Ben Davis

26,936 views • 5 months ago

Moved some internal tools over to Railway mostly just cause I thought it would be fun and wanted a long running server, did not expect it to be this absurdly fast This vid is not sped up at all, it's just a SvelteKit project, using bun, & mysql over railway's internal network
0:20
davis7's profile picture

Moved some internal tools over to Railway mostly just cause I thought it would be fun and wanted a long running server, did not expect it to be this absurdly fast This vid is not sped up at all, it's just a SvelteKit project, using bun, & mysql over railway's internal network

Ben Davis

36,356 views • 10 months ago

0xSero helped me setup local models properly and I uh, had no idea these things had gotten this good Are they frontier level? No, but considering this is running on just my 5090 it's remarkably capable First tests on a couple of programming tasks and the qwen 3.6-27b model with no reasoning feels about on par with something like sonnet 4-ish, probably better it's really impressive But also setting up local models isn't easy, I don't know nearly enough to talk much about it yet other than you need to know what you're doing to have a good experience. The out of the box stuff is not nearly as good as setting it up correctly
0:45
davis7's profile picture

0xSero helped me setup local models properly and I uh, had no idea these things had gotten this good Are they frontier level? No, but considering this is running on just my 5090 it's remarkably capable First tests on a couple of programming tasks and the qwen 3.6-27b model with no reasoning feels about on par with something like sonnet 4-ish, probably better it's really impressive But also setting up local models isn't easy, I don't know nearly enough to talk much about it yet other than you need to know what you're doing to have a good experience. The out of the box stuff is not nearly as good as setting it up correctly

Ben Davis

14,490 views • 3 months ago

I re-subbed to Claude Code to test out Opus 4.8 It's both better and worse than I expected tbh. Claude models have some really weird behaviors/hallucinations and seem to be getting slower That said, workflows are dope and the model is still very good...
25:55
davis7's profile picture

I re-subbed to Claude Code to test out Opus 4.8 It's both better and worse than I expected tbh. Claude models have some really weird behaviors/hallucinations and seem to be getting slower That said, workflows are dope and the model is still very good...

Ben Davis

10,936 views • 3 months ago

No more content to load