Haha this is surprisingly and exactly how I use claude as well. Quite fascinating that we independently discovered the same workflow.
I maintain two directories: "docs/proposals" (for the research md files) and "docs/plans" (for the planning md files). For complex research files, I typically break them down into multiple planning md files so claude can implement one at a time.
A small difference in my workflow is that I use subagents during implementation to avoid context from filling up quickly.
Same, I formalized a similar workflow for my team (oriented around feature requirement docs), I am thinking about fully productizing it and am looking to for feedback - https://acai.sh
Even if the product doesn’t resonate I think I’ve stumbled on some ideas you might find useful^
I do think spec-driven development is where this all goes. Still making up my mind though.
This is basically long-lived specs that are used as tests to check that the product still adheres to the original idea that you wanted to implement, right?
This inspired me to finally write good old playwright tests for my website :).
Spec-driven looks very much like what the author describes. He may have some tweaks of his own but they could just as well be coded into the artifacts that something like OpenSpec produces.
I like this take. In other words, AI is another abstraction layer. The same way Java is an abstraction layer for C/Assembly/binary (i.e. pick whatever base layer you want). This also implies that the software development industry has to change inevitably. For example, maybe craftsmanship is a mentality of the past.
Great post indeed! I totally agree that embeddings are underrated. I feel like the "information retrieval/discovery" world is stuck using spears (i.e., term/keyword-based discovery) instead of embracing the modern tools (i.e., semantic-based discovery).
The other day I found myself trying to figure out some common themes across a bunch of comments I was looking at. I felt lazy to go through all of them so I turned my attention to the "Sentence Transformers" lib. I converted each comment into a vector embedding, applied k-means clustering on these embeddings, then gave each cluster to ChatGPT to summarize the corresponding comments. I have to admit, it was fun doing this and saved me lots of time!
I am loving this idea already. My Twitter timeline is a clusterfuck of political contents, which is exhausting. All I want is to indulge myself in entertaining posts and nothing more. I use Twitter as a way to relax and let my mind breathe. So this is wonderful to see!
Any plans to support Firebox on Android? I heavily use Twitter on my phone so would be sweet to install the extension on my phone.
Twitter already has a built-in solutions to it, and has always had it. Just click on the 3-dot button on the kind of posts you don't want to see, and select "Not interested". And engage more with the kind of posts that you would like to have, and stop liking or reposting political stuff no matter how much you agree with it. The algorithms merely reflect your own patterns and preferences, if you find that exhausting it isn't the algorithm's fault.
My Twitter timeline is close to pristine, and on top of it Twitter has even started to promote posts from smaller accounts on the same topics so that I get even better exposure in things I like. It is awesome. You just need to have some hygiene.
Honestly I use those features extensively and it never seems to work properly, there's always a few posts that I dislike and shouldn't have been on my feed (like politics, news etc...); which is why I built mindfirewall in the first place.
Had the same thoughts which is why I built mindfirewall in the first place, glad to hear this is something you'd like.
Firefox is super easy to support from a chrome extension so will add that next (just didn't have the time yet).
If you want to use mindfirewall on your
Android: Download Kiwi browser
iphone: Download Orion browser
These should let you use mindfireall on your phone, although let me know if you run into any issues. My twitter is @ navigatoragi and my email is nav@mindfw.com
I don't think using the verbatim git history as a change log is a good idea.
But if you insist on using the git history for change log, why not use an LLM? You could feed git history to chatgpt and get a nice user-friendly change log.
I bet a lot of commenters here don't understand the extent to which this ruling will apply.
The ruling affects cases where statutes do not explicitly grant interpretive authority to the enforcing agency. When a statute clearly delegates interpretive authority to the enforcing agency, the agency's interpretations will definitely prevail in the courts.
So this ruling is not as terrible as most comments think. If anything, it's very rational. If a law explicitly applies to X, you don't want the executive branch applying it to Y and Z. That's how you create a rogue executive branch (even though you may be well intentioned).
I maintain two directories: "docs/proposals" (for the research md files) and "docs/plans" (for the planning md files). For complex research files, I typically break them down into multiple planning md files so claude can implement one at a time.
A small difference in my workflow is that I use subagents during implementation to avoid context from filling up quickly.