The new build, test loop for automating a repetitive task that has non-deterministic features?
Build:
Babysitting Claude Code, asking it to write up the task as a skill for future use.
Test:
Ask Claude to one-shot via the skill.
Repeat:
Update skill until the Claude can one-shot without guidance.
Urgh.
Maybe someone can put together a labeler for these folks/replies. 🤔
📍 El Cocinero
Restaurant · Singapore, SG
★★★★★
Delicious tacos
#BeaconBits
Was just thinking this as I was trying out Go and Python interop and coming across: www.reddit.com/r/rust/comme...
This fungus reminds me of Project Hail Mary, a book, and also the upcoming movie. I wonder if the timing of this article is correlated with the movie..
Can this be combined with strudel.cc for playback of sets? 🤔
Lol I thought I was so original with sirh.cc 😆
Like this pattern for networking 📌
Anyone has been using @getgrist.bsky.social ? How do you like it?
It looks like what I've been looking for with Coda.
Anyone use zero-instrumentation profiling tools for Python? How’s your experience with them?
E.g. perforator.tech
Dug into the source, looks like exclusions are managed here github.com/withastro/as...
Will contribute an excludeLangs option when I get a chance.
Ran into this issue.
Syntax highlighting breaks the default way of rendering it via code blocks.
I haven’t found a way to turn off syntax highlighting only for mermaid while leaving it on for other languages.
This is frustrating given I’d think this is a common use case since they support GFM.
What is the best solution for using mermaid diagrams in Astro? I've failed a couple of times at implementing it now.
@webpro.nl @ssp.sh @cassidoo.co in case you’ve taken a stab at mermaid rendering on your blog, curious what you’ve came up with.
I haven’t dug into the syntax highlighting internals but perhaps I should.
…and GFM (GitHub Flavored Markdown) supports mermaid rendering.
Ran into this issue.
Syntax highlighting breaks the default way of rendering it via code blocks.
I haven’t found a way to turn off syntax highlighting only for mermaid while leaving it on for other languages.
This is frustrating given I’d think this is a common use case since they support GFM.
Was wondering who else heard that…
Future LinkedIn skill:
* Able to persuade diverse set of LLM models.
Debugging an error coming from an AI agent:
`Expecting value: line 1 column 1 (char 0)`
As I step through the code, I see why.
The LLM (llama3.2:1b) is refusing to provide a response, as it thinks it could be used maliciously.
Guess the next step is to persuade this model that I'm 😇...
Interesting:
Meta created React Native. It’s used (with components at least) in their flagship apps: Facebook (iOS, Android), Instagram (Meta Quest), Messenger (desktop).
Google created Flutter. And yet none of their flagship apps use it (Gmail, YouTube, Maps, Calendar).
I meant that typically two modes (mocks, real endpoints) are kept available in the test code.
If the test strictly requires mocks and will never run without mocks (i.e. a mode that records new mocks such as when API endpoints have updated) then I typically regard it less of an integration test.
It’s okay to do so but there’s still a way to run it without the mocked APIs to hit real endpoints.
It’s not that they must rely on mocks being available.
Book that comes to mind is: Non-violent communication.
Not exactly about empathy directly but a tool to help with it.
Used ChatGPT as an editor, and published it. Done is better than perfect, heh.
Ah, just saw this. You beat me to it.
