Last 2 weeks I was using it more or less all day on Opus running skills to write PRDs and then code and tests to solve the PRDs, never hit the session limit.
Last 2 days I hit the cap in about an hour of kicking off my skills workflow.
On the paid enterprise team plan this is really bad.
If you have mapped the research space and understand it very well then what does it matter how you got there?
For example when I was doing my master's degree in the UK, I realised my research topic had about 10 people around the world working on similar problems and I had read all of their papers.
I could trace the early research and seminal works in the field from the 1960s until now and knew by name (and often email correspondence) the 10 or so researchers working in the hyper-niche space.
If you can get to some kind of intimate understanding about your body of research the method is not really relevant. But take the LLM away can you still hold a conversation with an expert? LLM cannot read the papers for you.
Frankly it's a total non-issue because you now have your train ticket on a mobile app on a mobile device with synchronized internet time anyway. Most people are wearing mechanical watches today as jewellery and a physical piece of craftsmanship that's nice to keep time that's directionally correct, but if it's out by 30seconds it's no big deal and you just reset it against the time on your phone and get on with your day.
I still use paper tickets and many others do too! At any rate though, I agree. It doesn't matter to me that my mechanical watch is often out by 30 seconds. I don't mind checking and adjusting it every now and then so it's never off by ±5 minutes where it would cause problems for me.
I’m not saying you’re wrong but as a senior PM, the engineers I work with see about 10-20% of what I actually do in a week, so in general engineers are not a good judge of the utility of product management.
You're right that this isn't some groundbreaking revelation. If you're using AI enough to be feeling it, you're feeling/seeing what they're talking about. The purpose of a paper/retreat like this it get it all together and written down on paper, then to disseminate it to the wider world. I think the paper does a good job of collecting info that isn't wrong, and which has enough info to help guide folks making decisions.
Mainly because Martin Fowler is part of their C suite
I agree that it's marketing material, but that doesn't instantly make it garbage. I've been reading their quarterly Thoughtworks Radar for a while now and it's clearly put together by people who understand the industry.
Sigh. Nobody is ever going to be happy. Would saying it came from a rando Reddit user be better?
They at least put the effort into having the retreat and putting this together. Would other consultancies (who we know little about) have of done the same?
Have you spoken to https://www.citizensadvice.org.uk/ you should be able to get some help and advocacy. And they will help you find accomodation and apply for the benefits you are entitled to.
You're obviously smart and probably present as high functioning to someone working in the job centre who is not motivated to help you and frankly anything they give you, you will be massively underemployed.
---
I will say something feels off with the project itself just by looking at the stats.
Obviously an LLM is doing the heavy lifting to write 1.3m lines in 4 months, but 3,900+ tests seems an order of magnitude too small to me.
What LLM are you using to generate this much code? Priority 1 right now is to get accomodation, not pay for expensive LLM API costs or subscriptions!
But I have to be honest: I’ve been through Citizens Advice and the benefits system for years. CA is a signposting service that directs you to the same services that have already rejected you. PIP requires a formal diagnosis, which requires years of waiting lists and assessments.
I’ve been fighting this system long before I ended up in a tent. The post explains that the state won’t help, and that’s not an exaggeration or an oversight on my part. I’ve tried.
On the test count: fair point. The 3,900+ is the core workspace (50 crates).
Satellite projects have their own test suites not included in that number. Coverage could be better. That’s part of what stability buys me.
On the LLMs: Claude Code for complex Rust architecture. Codex for pattern-heavy work. Gemini for docs and broad context. Qwen for parallel workloads. All running simultaneously with strict file ownership and pre-defined interface contracts.
But to be direct about what I’m actually looking for: I don’t want benefits. I don’t want PIP. I don’t want permanent support from the state. I want to use my brain, build complex systems, and turn AI research into reality.
What I need is 3 months of runway so I can do exactly that.
> What I need is 3 months of runway so I can do exactly that.
I think you need to be more realistic about your priorities right now, you need a job, any job, and a roof over your head mate.
The OS and runway for a start up can wait. This is not an investable proposition for anyone, what happens after 3 months of runway on a vibe coded OS?
Where's the go to market plan, where's the viable business here? I'm not saying none of that exists but you need to put yourself on more stable footing before working on it, otherwise it will remain a pipe dream.
Please show me a vibe coded OS containing over 20 separate projects all merging together.
A from scratch kernel, from scratch hypervisor, from scratch everything.
Theres a big difference between systema architecture with AI tooling and vibe coding a little app.
This is not a vibe coded application.
I'm a 40 year old man. If I had the option to go get a job I would.
What would I do with my dog while I'm at a job? Do I leave him in the tent while I go off for 8 hours or more each day?
I'm guessing you don't fully grasp the point I'm aiming for here.
I'm hoping someone will see this and understand that I'm better than a dead end job. My brain is very unique and has the ability to do things others can't.
Honestly, if I could just go out and get a job and a place to live in would not be posting things like this.
Go beg with your dog until you make enough for a case of water and supplies (marker / board). You sell your water for a dollar until you have made enough for a cooler, ice and more water. You keep selling water daily and move locations until you are making $100 a day. You take the money get an apartment and keep saving.
You take your savings and go to every rust conference / local meetup you can afford to travel to. You write blogs, make connections in the rust community.
If you are lucky someone might use it and write a nice tweet which will get more people to notice. One day someone at a large company will use it and pay you for support.
or
Bring the dog with you job hunting do the 35 hours and get support.
I'm not going to beg on the streets tbh. I'm not willing to stoop down to begging on the streets.
I'm not looking to sell water.
I have to be honest, i don't think you understand how real life works.
Your suggestion is to to sell water to make a few pennies and that will get me £2000 for a place to live and allow me to travel the world attending rust conventions.
I'm looking for help, not silly ideas that you saw in a movie once.
Besides anything it's literally illegal to sell food based products (drinks included) here in the UK without a proper license and it could literally land me in court for tax avoidance and all sorts.
Maybe in the US things are different but here in the UK we can't just sit at the side tof he road selling water - it's 5°C out, people don't by water from beggars in the UK.
10-12 parallel AI coding sessions running simultaneously.
1.3M lines over 4 months at ~18 hours a day works out to roughly 600 lines per hour total, or about 60 lines per hour per session. One line per minute per session.
AI writing makes me irrationally angry, when I use AI and tell it to avoid everything from https://en.wikipedia.org/wiki/Wikipedia:Signs_of_AI_writing and it does improve things. Banning em-dashes, and it's not x it's y false equivalences.
And I use it for myself and what I send to others is 99% written by me.
For people who treat writing, especially business writing as a craft to communicate ideas, seeing AI slop is just like nails on chalk board.
I genuinely care about this stuff. Thoughtlessly blurting out pages and pages of vanilla unedited LLM output seems disrespectful to the reader.
As writer you're saying: I didn't care enough to craft my message personally, here read this generated content I haven't even seriously edited.
And for the reader it's saying the same: This guy sent me a document to read, I need to sift through it to figure out if there's any actual merit or novel ideas or actional information here.
I think AI tooling has actually reminded us how useful the terminal really is. running claude code, running formatters and linters and test suites, managing incremental work with git etc. Are all vital in the modern way we write code. I don't see how you can put a glossy UI on top of that and maintain any fine-grained control. Since we're already giving up a LOT of control by having agents write our code, we need to retain some control and certainty over the quality control and review process of the outputs.
- I'm seeing prototypes escape Figma and live as code for a faster/closer demo experience for product managers.
reply