Haha right on all accounts - yeah been stuck on old skool and avante garde electronica ever since, so lucky we got to live through that era, wouldn’t change a thing, even the resulting tinitus! Definitely taking you up on that offer, there’s loads of nights I’ve wanted to go to but my mates just aren’t into it.
Yeah git is my firewall for the auto-changes from Claude Code, it gets creative, needs to be reigned in quite often. Took the leap and went Max but only for one month as I’d got a lot to get done and kept hitting the limits even with my efficient prompts, squeezing the most out of sonnet, but now all bets are off and I say bring on the limits! Got a ton of things queued up for when Fable comes back, that thing was incredible, if Opus was a potato gun Fable felt like a laser powered rifle.
Going to be an exciting week, got the carrier board turning up, nearly ready to launch the pre-launch site for my kickstarter, will probably evaporate but I don’t care, we HAVE to try these things!
At risk of this turning into a DM thread whereabouts are you in London? I grew up in Enfield, mum still lives there, but am in Stevenage because of kids!
That’s interesting, I’ve felt 5.5 was token hungry, I’ve never been able to get anything out of it completely before the shutters come down - I’ve access to ChatGPT and Claude at work (use Claude Max for personal), although one thing ChatGPT does really well is giving an occasional limit reset token, and the thing I like a lot is to use a pending message to “steer” the investigation, as I’m always worried about stopping Claude mid-thought!
Apparently (and take this with a pinch of salt) Fable will be back in a couple of weeks - hopefully back on the non-API token based subscriptions for a short time, I have things queued up for it! I know ALL of this stuff is massively opinion-related, but Fable felt significantly better than what’s currently available.
How are you using AI in your work kindo9? For me it’s infrastructure and IaaC related at work, and Pi related projects at home - with Claude Code (Max) on Auto mode with solid context memory this opens up so much potential. I love hearing about other’s experiences and what works for them too!
(apologies Clive, et al, maybe this should be another thread )
Myself I didn’t get on with Claude, I think they were having issues at the time; my subscription, I asked one question and it used all my limits !
I don’t tend to use the CLI versions, I still like the IDE, I use VScode with Codex Extension.
Getting it analyze massive csv logic captures has been very good in analyzing driver frames. I just add a references folder to my project, and dump big files into it and it can reference them.
It amazes me more everyday, I was adding features to a compiled Android app the other day, it was writing the Smali decompiled code.
GPT 5.5 has been super accurate from my experience, not needed to shop around again atm. Hopefully we get the new models, it seems we are reaching government only models now.
Codex is cool, I like the way it interacts also, and yeah they’re a much of a muchness really, this last window of Fable 5 usage I really hammered it and while it was good, it didn’t blow me away like some YT videos suggest. I am only running on Max x5 for a month, but still have managed to run out of my weekly window in a couple of days but been hitting it hard. I know exactly what you mean, it’s incredible - I’ve been programming for decades and some of things I’m doing I just couldn’t get done in a short timeframe, it’s massively accelerated everything I do, what a time to be alive!
)