It's Claude Time again! Today @rendle.dev and I are joined by @richcampbell.bsky.social - let's see if Claude Code can build user interfaces? Can it understand sketches and scribbles?
Tune in: twitch.tv/dylanbeattie
(or youtube.com/dylanbeattie if Twitch isn't your jam)
On my next refresh, I'm going to experiment with resuming the prior .NET engineer session for the fixes, rather than starting a new session. I'm not sure if a clean context is better or worse i that scenario.
This ends up burning usage even with Sonnet 4.6 which I could argue should have been reduced if the first step actually did everything. Granted this mimics real life, where a fresh look finds issues, but now this has a direct token cost for me. I'm running on Claude Pro so have limited tokens. 3/4
Since QA often finds gaps, even things simply not done, I have it build a file of blockers and recommendations to pass back to the .NET engineer. Then that churns to fix its prior mistakes. Then I pass to QA again and if that's happy, do my manual code review. 2/4
My current frustration with my AI agent workflow experiment, specifically Claude Code is how quickly I burn my 5-hr usage on basically one small feature. I run in a #dotnet engineer persona to apply the OpenSpec change, then switch to a QA persona to review. 1/4
Sounds cool. I'm mostly manually running 1 to 3 CLI agent sessions concurrently but at some stage want to look at sub-agents and auto-orchestrated workflows. Still getting the basics right for the workflows that work for me.
I'll continue to refine a workflow and at some point, hopefully put some blogs and/or videos together to share how I'm leveraging AI in this way. #ai #claudecode #cursor #copilot
It's been an interesting experiment on how to leverage AI to do things I have on the backburner but never find time to do. I have tons of (years old) reserved domains I could build out this way. A low touch, but structured workflow makes it theoretically possible for me to make progress.
I'm also refining my own workflow for tracking the work we're doing and automate a few things with my own commands/skills. So far I am getting something built. It's not been perfect but it's allowing me to iterate on an idea I'd never have found time to build from scratch.
I'm using the OpenSpec flow and different personas, to refine my idea into requirements, plan out the implementation and then build the features. I'm generally grabbing 30 mins a day to prompt the next steps, and a few mins to review it.
I'm currently on holiday and I've been experimenting with using a multi-agent workflow, in part based on the great ideas presented by @rendle.dev on their stream with @dylanbeatt.ie to build out an idea without eating into family time.
Yet more Claude talk and experimentation and such
www.youtube.com/watch?v=c2rQ...
Seems like @rendle.dev and my ad-hoc "wtf is Claude Code" series is getting some traction, which is heartening.
We'll be on Wed + Thu + Fri this week, starting around 2pm UK time, live on Twitch (twitch.tv/dylanbeattie) and on YouTube (youtube.com/dylanbeattie). Tune in and chat - or catch up later.
Side table in a bedroom covered in medical supplies including infusion kits, needles, a syringe and HyQvia medication.
A bedroom with side table containing medical supplies for infusion, a sharps bin and a hospital grade pump next to the bed.
Today was infusion day for me, which comes once every three weeks. I get to top myself up with fresh new antibodies. It requires turning my bedroom temporarily into a hospital! I've gotten used to the process after 6 months of doing it. A bit sore tonight and exhausted. #immunocompromised #cvid
Watched both recordings so far and it's been great. I'm at that point of, I need to learn this stuff and make more use of the tools now, so having guidance from two people whose opinions I respect and trust is very helpful! Looking forward to part 3.
I got a request passed along from a potential customer last week asking for a workaround to run our library on .NET Framework 2.0!!
Sage advice!
I'll likely aim to assess both over time. I'm leaning toward Claude first as that's the one I've had the most positive feedback on from colleagues. But great to have your opinion too as it's highly valued!
No, I've used it a little with the free plan only.
Herein lies the problem! So much to learn about and try out and next week there will be some new thing getting hype. I wanted to try Claude next as its the one I hear best feedback on. How do these compare in your experience?
I'm watching and reading a lot of content on using AI coding agents to start to build up my knowledge. At some stage, perhaps I'll distill it into blog posts of my own. With things moving so fast, im not sure of the value in that until things stabilise. WDYT?
I'm skill unsure of the sustainability with regards to companies eventually needing to charge a huge amount more to cover costs. Where we'll end up exactly is unclear. But understanding the fundamentals, tricks and "best practices" sooner rather than later feels sensible.
In some form, they will be here to stay and I think its important not to get left behind. We still need humans, but more and more tasks will be delegated to AI. Finding out how these tools fit into my workload and how to leverage them best is now important.
We're at the point where people whose opinion I respect, including colleagues and those in the #dotnet community, are providing positive feedback and examples, making it hard to ignore the progress. Things are moving fast, but I think now is the time to start learning these tools.
As someone who was cautious and skeptical of AI coding agents, I'm starting to accept that these tools are hugely improved compared to where they were even 6 months ago. I made good progress on some changes fully in @github.com Copilot CLI yesterday. Next week I plan to try Claude Code.
Three year old sitting in the captains seat on board an A320 EasyJet flight.
I'm not sure the pilot for our flight is fully qualified!
Blogged: "Encrypting Properties with System.Text.Json and a TypeInfoResolver Modifier (Part 2)"
www.stevejgordon.co.uk/encrypting-p...
Continuing our journey by adding a component that can encrypt and decrypt data using AES-GCM.
#dotnet
Submitted 2 security issues to NuGet and GitHub last year, finally heard back that they are low priority and not a big deal.
I did manage to publish this package to NuGet. A bunch of unicode in there, but it looks genuine, right? Nobody would discover this via search because of the unicode.
Just blogged, live from @ndcconferences.com #ndclondon, "The Grand Mystery of the Missing 18 Bytes". An exploration of the implementation details affecting different values from benchmark allocation data when compared to memory profiler data in #dotnet.
www.stevejgordon.co.uk/the-grand-my...
You gotta love how CoPilot will confidently lie to you. Gave it compiling code to, asked it to find and fix a small bug. It found the bug (correctly) and fixed it, then builds and it fails. It locates a malformed byte literal and states "There's a syntax error in the file (not related to my change)"
