@dracani
iAccount based inUnited Kingdom
About this account
- Account based in
- United Kingdom
- Connected via
- United Kingdom Android App
Account-level information from X, not a live location or the device used for a specific post.
π Links to all my socials, the podcast, newsletter, YouTube, courses, blog, and more. can be found on my personal website: https://nitter.cf/t.co/2EKxnr4ZeO
Oxfordshire, UK
Joined November 2008
- Tweets17K
- Following2.4K
- Followers3.9K
- Likes15.7K
I sent the short update email this morning, and getting lots of really nice complimentary messages back about the newsletter π It's really good to hear that subscribers are getting value from it.
I've moved my newsletter to @buttondown.
You can subscribe here if you're not already subscribed: danclarke.com/newsletter/
If you are already subscribed, there's nothing to do - I'll send a short email on Monday about the switch.
I do like the way @buttondown sends me an email with a different fun celebration giphy each time someone subscribes! π
I've moved my newsletter to @buttondown.
You can subscribe here if you're not already subscribed: danclarke.com/newsletter/
If you are already subscribed, there's nothing to do - I'll send a short email on Monday about the switch.
I've moved my newsletter to @buttondown.
You can subscribe here if you're not already subscribed: danclarke.com/newsletter/
If you are already subscribed, there's nothing to do - I'll send a short email on Monday about the switch.
π€ Made with AI
Claude does look after me! I've been meaning to go to bed, but I keep on procrastinating in Claude Code. I'd best do as I'm told! π€£
π° Newsletter 047 is out.
Anthropic's Opus 5.5 is out and getting really good press, which is great after the hate Opus 5 was receiving! Claude is back! π
And OpenAI also released GPT-6 Sol and Luna on the same day.
danclarke.beehiiv.com/p/newsβ¦
Oh, just noticed that slash commands now work anywhere in the prompt! π₯³
Going in the other direction now - something that I wish Claude Code did that GHCP CLI does, is allow slash-command completion anywhere in the prompt. CC only seems to show the dropdown if the prompt is empty. Sounds trivial, but actually being able to do "/" mid-prompt is really useful. @bcherny
Opus 5.5 is the result of your feedback.
It communicates clearly, it's cheaper per token than Opus 5.0 with the intelligence of Fable 5.1, itβs very token efficient and works across every effort level.
We're also increasing 5h rate limits & giving you a banked reset.
Dan Clarke π retweeted
If you are into Claude Code, this is my chat with @dracan. Listen till the end. It has a couple of super useful tips.
ποΈ New episode has just dropped! @gsferreira is back on the podcast for the first of two episodes, and this time we're chatting about Claude Code, including our favourite features and tips! #UnhandledException
unhandledexceptionpodcast.coβ¦
I do find that Astra needs to be explicitly told what to do next a lot more than Claude models do.
Claude models normally suggest the obvious next thing and ask me if I want it to go ahead - and quite often all it requires from me is a "yes" response.
With Astra, I have to explicitly tell it what the next thing is a lot more.
This fixed it in my current conversation:
"For the rest of this conversation, can you get a bit better at prompting me with what's next, rather than making me second-guess?"
It responded with:
"Yes. I'll end with a clear recommended next step and say whether I'm waiting for your instruction."
And it's now working much more like Claude does in that respect.
I think I'll just put something similar in my global instructions.
Really enjoyed listening to this conversation between @GergelyOrosz and @mattpocockuk
Definitely worth a listen.
newsletter.pragmaticengineerβ¦
ποΈ New episode has just dropped! @gsferreira is back on the podcast for the first of two episodes, and this time we're chatting about Claude Code, including our favourite features and tips! #UnhandledException
unhandledexceptionpodcast.coβ¦
Dan Clarke π retweeted
Thinking about creating a /pr skill. You invoke it to create a PR that:
- Makes human review as fast and as simple as possible
- Takes the best parts of /show-me by @dexhorthy
- Uses your domain language (from /grill-with-docs)
- Shows evidence that the change works (images, test output etc)
- Assesses merge risk (is it a one-way door, or a two-way door?)
IMO every model/harness I've seen creates garbage PR bodies - needs a skill here.
WDYT?
Podcast editing is by far what takes the most of my time as part of hosting a podcast. Whilst I would like to just outsource it - the amount of editing I actually do on an episode would cost a small fortune to pay someone else to do it. So I've starting using AI to create a bespoke podcast editor so I can really fine-tune and automate my workflow as tightly as possible. I've used @mattpocockuk `/wayfinder` skill quite a bit recently and it's been amazing for building various different projects - so I think this will be a really good fit for this project too, and I know Matt's done similar for his own bespoke video editor. I will keep you all posted on how the podcast editor build goes! π€
π° Newsletter 046 is out.
Anthropic has rebuilt Projects, starting in Claude Code - you describe the work in one conversation and Claude splits it across parallel cloud threads.
danclarke.beehiiv.com/p/newsβ¦