I wrote a new blog post about the culture of hard work in tech, perfectionism, self-compassion and my dreams for the tech industry. You can read it here:

Self compassion and the disposable engineer dlants.me/self-compass...

@udini.bsky.social really enjoying the book! For me it's been a nice reminder to resume my efforts to let go of chasing measurable strength and chasing fatigue. Leaving the gym when you're not tired is tough!

Notes – 12:08 Wed 10 Sept 2025 Notes – 12:08 Wed 10 Sept 2025

noted: have you seen what Bun are up to? Yikes... 💀

dbushell.com/notes/2025-0...

The Definitive Diet Setup Guide: How to Build and Adjust a Smart Nutrition Plan This guide covers everything you need to know to set up a well-structured diet that supports your performance and body composition goals.

strongerbyscience is great for no-nonsense, evidence-backed nutrition advice. Some of it angles towards "getting ripped" but it is generally grounded in science and approaches it with a great 80/20 mentality. You might want to start from this guide: www.strongerbyscience.com/diet/

Pdf pager by dlants · Pull Request #184 · dlants/magenta.nvim The previous solution of converting pdfs to text wasn't really working for me - too much info got corrupted in the translation. I think this sets magenta up to work well with semantic search, s...

One of my dreams for my #nvim / #neovim ai plugin is to use it as a writing / research assistant. I took a small step forward on that today by improving the way it handles reading pdf files github.com/dlants/magen...

They don't :(

So the best way to get access to gpt-5 for this purpose is to set up a script that burns through $50 of tokens to get you to tier 2. That's absolutely bonkers! Just let me preload my account with $50 of credit to get access to the higher tier!

30K TPM in tier 1 basically prevents you from being able to use a coding assistant. And there's no way to pay your way to a higher tier, you have to get there by organically spending $50

I develop a coding assistant and OpenAI feels really hostile towards developers. Their sdk is a mess (does not adhere to jsonschema, is poorly documented).

Recently I tried to run some experiments on gpt-5 and immediately ran into their rate limits / tiering system.

Open source tools should just let you define a regex for finding context files. (Like `autoContext` in magenta github.com/dlants/magen...)

- I added authentication via anthropic's max and pro plans, which lets you use tokens the same way as claude code (thanks to the opencode folks for their open-source implementation for how to do this). This is much cheaper relative to using the pay-per-token API for heavier users.

- I renamed "@compact" to "@fork", and also changed it from a forced tool use to just a regular tool. This means the agent can think before using it, improving the quality of the summary, and also this now takes advantage of the conversation cache, making these tool calls a lot faster and cheaper

GitHub - dlants/magenta.nvim: A tool-use-focused LLM plugin for neovim. A tool-use-focused LLM plugin for neovim. Contribute to dlants/magenta.nvim development by creating an account on GitHub.

Some updates for my #neovim #nvim ai plugin github.com/dlants/magen...

- I improved the turn-taking. You can now interrupt the agent by sending additional messages or use the @async prefix to your message to enqueue it to be sent at the next turn-taking opportunity

i tested it, it works

Claude max auth by dlants · Pull Request #175 · dlants/magenta.nvim

🎉 github.com/dlants/magen...

I can now have multiple ghostty windows open, one per tab, and use a command shortcut to quickly switch between them :tada:

Now I just need to figure out how to make my ghostty sessions persistent, and do some tweaking around how tabs are numbered, and I'll basically be there.

this uses hammerspoon to launch a ghostty terminal. There it runs a script that uses hammerspoon's cli to query the open windows and tabs (using the stuff that ghostty exposes for a11y). It then feeds it into fzf, and when fzf returns a selection, uses hs again to switch to that window/tab.

After a night of hackery, I was able to replicate something workable using #hammerspoon and ghostty's accessibility features! github.com/ghostty-org/...

After watching @mitchellh.com talk about this, I was inspired to try eliminating tmux from my workflow and using #ghostty directly www.youtube.com/watch?v=o-qt...

One thing I missed was being able to have a tmux session per project, and my fzf session switcher github.com/dlants/dotfi...

a simple graph illustrating what the target click zones for d3 ticks are in red rectangles

ok this was actually pretty dope. I asked claude sonnet 4 (using my magenta plugin) what the actual clickable area of a "tick" was in d3. It created a demo html page, opened it using the playwright mcp server, and ran some experiments to give me a grounded answer.

I'm seeing a lot of commentary blaming the egregious data leak from the Tea dating safety app on vibe coding

I'm confident that, in this particular case, that's not what happened: the code at fault looks to have been written back in late 2023 simonwillison.net/2025/Jul/26/...

One cool thing is that you can use options to [append to the system prompt](github.com/dlants/magen...). I think by providing additional [examples of how you want it to behave](github.com/dlants/magen...), you can have it be a lot more useful for your specific use-case.

sofia (@sop@unstable.systems) Attached: 1 image A shitpost that's been brewing in my head for a while now

I also opted to have this be manually triggered (Shift-Ctrl-l by default in insert or normal mode). This is a lot [less distracting](unstable.systems/@sop/1148985...).

One thing that is interesting about this is that I wasn't able to get good results from smaller/faster models like claude haiku, so I opted to use a larger model instead. I found that the small delay of about a second was worth it for more consistent results.

- Finally, I show the find/replace in the buffer using [virtual text extmarks](github.com/dlants/magen...), applying a comment effect to the added sections, and a strikethrough to the removed sections

- I then feed that along with the [context surrounding the cursor](github.com/dlants/magen...), and a [system prompt](github.com/dlants/magen...) into an LLM, [forcing a tool use for a find/replace within the context window](github.com/dlants/magen...)

- I used an [lsp server](github.com/dlants/magen...) to track opened files and aggregate text changes to get a stream of diffs.

a short gif of a terminal demonstrating next-edit-prediction. You can see that there's a "completing..." virtual text when the completion is in flight. Then the proposed edit is shown by overlaying strikethroughs and new text in the buffer.

I added next-edit-prediction to my #neovim #nvim plugin, magenta.nvim

This was pretty interesting to implement! (thread)

GitHub - dlants/magenta.nvim: A tool-use-focused LLM plugin for neovim. A tool-use-focused LLM plugin for neovim. Contribute to dlants/magenta.nvim development by creating an account on GitHub.

if anyone wants to give me $5000, there's a sponsor button on the github page for the plugin <3 github.com/dlants/magen...

Dunno if you're a neovim / vim enthusiast, but if you like reviewing coding tools, you should check out github.com/dlants/magen... . I wrote about some of its design philosophy here dlants.me/ai-whiplash....