@NoBSRecruiter I have recently upgraded to the $200 plan.
I blame all the recent resets really, I swear they'e contributed to my higher usage.
They're both a blessing and a curse!
I did this a couple of months ago - much of my backend experience was with monoliths. Where microservices were used in previous roles, we had really basic polling rather than modelling them as durable jobs.
I made an app that collects all your bookings/accomodations in a single place, scraping your emails for details from Airbnb, Airlines etc.
I had an Azure Function worker and used Azure Service Bus to handle the AI work asynchronously. Easier said than done, not easy at all!
@ForwardEditor The tl;dr of voice mode for me:
Is it good for me, in terms of improving my communication skills? Undoubtedly.
Do I want to do it / can I be arsed? Nope...
I would like something similar maybe - on my Phone I am more or less limited to using Codex (assuming my laptop is still open).
My workflow on my phone is really:
1) Ask ChatGPT stuff (thats not linked to Codex)
2) (Very occasionally) use Codex remote but that relies on my laptop being open.
I'm sure theres some interesting ways to use AI through my phone beyond these fairly basic ways.
Then again, I think maybe its for the best, sometimes its good to get a break from the LLM's...
@DavidKPiano These days I just modify the personalization thing so it does it in a single step without me asking.
If I really care about whats being committed this feature in VSC is great. Would have been useful a while ago pre-AI if I bothered to research it
I agree with this a lot:
After spending 30+ years in front of a computer converting thoughts=>characters, its not so easy to go from thoughts=>voice=>characters.
It almost adds an intermediate step.
I do think it would benefit me to speak my thoughts out loud though using Whipr - I think its a no-brainer if you want to improve your communication skills at work or with clients for instance.
I've experimented with doing everything through an agent the last couple of days - basically as a 'Life' pair programmer.
It manages Todo lists, project statuses, things I need to remember, journeys etc. I think theres a backend involved - who knows, I have delegated my intelligence entirely to an AI model. I
Perhaps this is something people already do. I am a boomer, so maybe so.
My thoughts:
+ It is quite reassuring to just run everything through an AI.
- It is still too damn slow for quick things. Latency of 30s is just not great. I think of the human brain a bit like an Azure Function - it has a cold start and after 30s it takes time for me to warm up.
- I am still paranoid about turning more and more into a machine the more I embrace AI a.k.a. Darth Vader style. More machine than man, as a colleague once told me a few years ago
@johnlindquist I gave up on Cursor back in February as the costs were extortionate, and the UI kept changing.
I'll give them another go once the dust settles from the acquisition.
I really didn't like the Cursor Glass appearance at all, so hopefully they've made some improvements to it.
@johnlindquist I did a technical test last night, answering about 50 of these kind of questions.
I gave up after about 20 of them - loads of weird JavaScript cryptic crosswords I don't care to solve anymore!
This is the problem with Skills - once you've accumulated enough you really have no idea which ones are actually effective (apart from @Aaronontheweb's of course...)
So I just hoard and carryon
Chrome has seemed to have nerfed AI-browser interaction on certain pages e.g. the Chrome Web Store Developer page. (Bit of a mouthful thatone)
I haven't bothered to check if I get the same issue on Google Cloud.
Wonder if more coming soon...
That was awesome!!
I think I struggled a bit being paranoid about being a bit slow to understand the problem. I reckon if he went away for 10minutes and came back I would have been fine.
Sometimes it takes me a while, and it doesn't help if the interviewer is stone silent throughout..
I think you would have gotten the job!
Remote EU 90k GBP Dubai hours if thats your thing but anyway enjoyed reading that Matt!
Had an interesting job interview a month ago.
In the technical round, I was encouraged to use AI as much as I want. Interviewer a CTO who's exited a few companies successfully and a sharp guy
I was really expecting to have to knock up a website in an hour, but it was actually a LegalTech domain problem.
Task was build a system to check if a lawyer can work on a case and no conflict of interests
Do you:
1) Literally ask Claude Code to scan the interview pack, and ask Claude Code what clarifying questions to ask the interviewer?
I.e. Just talk with Claude
2) Attempt to actually read the problem yourself, understand it before touching an LLM
Genuinely curious !
I use it a lot for doing quick end to end tests of the whole application, and GCloud stuff that isn't really doable via terraform.
My biggest worry is token costs - I'd like to experiment more with it but rather cautious about blowing through a bunch of my plan doing more experimental stuff!
If you can't get it to work in Codex you can try OpenCode which allows usage of your Open ai subscription
Ctrl+v works on a Mac for me in ghostty. Hopefully it works for you in VSCode terminal - in the past I had some issues in VSC where some commands are "swallowed" so maybe worth a debug if alt+v doesn't work