Can you give me an idea of how much interaction would be $50-$100 per day? Like are you pretty constantly in a back and forth with CC? And if you wouldn’t mind, any chance you can give me an idea of productivity gains pre/post LLM?
Yes, a lot of usage, I’d guess top 10% among my peers. I do 6-10hrs of constant iterating across mid-size codebases of 750k tokens. CC is set to use Opus by default, which further drives up costs.
Estimating productivity gains is a flame war I don’t want to start, but as a signal: if the CC Max plan goes up 10x in price, I’m still keeping my subscription.
I maintain top-tier subscription to every frontier service (~$1k/mo) and throughout the week spend multiple hours with each of Cursor, Amp, Augment, Windsurf, Codex CLI, Gemini CLI, but keep on defaulting to Claude Code.
I am curious what kind of development you’re doing and where your projects fall on the fast iteration<->correctness curve (no judgment). I’ve used CC Pro for a few weeks now and I will keep it, it’s fantastically useful for some things, but it has wasted more of my time than it saved when I’ve experimented with giving it harder tasks.
It's interesting to work with a number of people using various models and interaction modes in slightly different capacities. I can see where the huge productivity gains are and can feel them, but the same is true for the opposite. I'm pretty sure I lost a full day or more trying to track down a build error because it was relatively trivial fpr someone to ask CC or something to refactor a ton of files, which it seems to have done a bit too eagerly. On the other hand, that refactor would have been super tedious, so maybe worth it?
Mostly to save money (I am retired) I mostly use Gemini APIs. I used to also use good open weight models on groq.com, but life is simpler just using Gemini.
Ultimately, my not using the best tools for my personal research projects has zero effect on the world but I am still very curious what elite developers with the best tools can accomplish, and what capability I am ‘leaving on the table.’
I’m a founder/CTO of an enterprise SaaS, and I code everything from data modeling, to algos, backend integrations, frontend architecture, UI widgets, etc. All in TypeScript, which is perfectly suited to LLMs because we can fit the types and repo map into context without loading all code.
As to “why”: I’ve been coding for 25 years, and LLMs is the first technology that has a non-linear impact on my output. It’s simultaneously moronic and jaw-dropping. I’m good at what I do (eg, merged fixes into Node) and Claude/o3 regularly finds material edge cases in my code that I was confident in. Then they add a test case (as per our style), write a fix, and update docs/examples within two minutes.
I love coding and the art&craft of software development. I’ve written millions of lines of revenue generating code, and made millions doing it. If someone forced me to stop using LLMs in my production process, I’d quit on the spot.
Why not self host: open source models are a generation behind SOTA. R1 is just not in the same league as the pro commercial models.
> If someone forced me to stop using LLMs in my production process, I’d quit on the spot.
Yup 100% agree. I’d rather try to convince them of the benefits than go back to what feels like an unnecessarily inefficient process of writing all code by hand again.
And I’ve got 25+ years of solid coding experience. Never going back.
> data modeling, to algos, backend integrations, frontend architecture, UI widgets, etc. All in TypeScript, which is perfectly suited to LLMs because we can fit the types and repo map into context without loading all code.
Which frameworks & libraries have you found work well in this (agentic) context? I feel much of the js lib. landscape does not do enough to enforce an easily-understood project structure that would "constrain" the architecture and force modularity. (I might have this bias from my many years of work with Rails that is highly opinionated in this regard).
When you say generation behind, can you give a sense of what that means in functionality per your current use? Slower/lower quality, it would take more iterations to get what you want?
Context rot. My use case is iterating over a large codebase which quickly grows context. All LLMs degrade with larger context sizes, well below their published limits, but pro models degrade the least. R1 gets confused relatively quickly, despite their published numbers.
I think Fiction LiveBench captures some of those differences via a standardized benchmark that spreads interconnected facts through an increasingly large context to see how models can continue connecting the dots (similar to how in codebases you often have related ideas spread across many files)
> I’ve written millions of lines of revenue generating code
This is a wild claim.
Approx 250 working days in a year. 25 years coding. Just one million lines would be phenom output, at 160 lines per day forever. Now you are claiming multiple millions? Come on.
It's impossible as an IC on a team, or working where a concept of "tickets" exists. It's unavoidable as a solo founder, whether you're building enterprise systems or expanding your vision. Some details -
1. Before wife&kids, every weekend I would learn a library or a concept by recreating it from scratch. Re-implementing jQuery, fetch API via XHR, Promises, barebones React, a basic web router, express + common middlewares, etc. Usually, at least 1,000 lines of code every weekend. That's 1M+ over 25 years.
2. My last product is currently 400k LOCs, 95% built by me over three years. I didn't one-shot it, so assuming 2-3x ongoing refactors, that's more than 1M LOCs written.
3. In my current product repo, GitHub says for the last 6 months I'm +120k,-80k. I code less than I used to, but even at this rate, it's safely 100k-250k per year (times 20 years).
4. Even in open source, there are examples like esbuild, which is a side project from one person (cofounder and architect of Figma). esbuild is currently at ~150k LOCs, and GitHub says his contributions were +600k,-400k.
5. LOCs are not the same. 10k lines of algorithms can take a month, but 10K of React widgets is like a week of work (on a greenfield project where you know exactly what you're building). These days, when a frontend developer says their most extensive UI codebase was 100k LOCs in an interview, I assume they haven't built a big UI thing.
So yes, if the reference point is "how many sprint tickets is that", it seems impossible. If the reference point is "a creative outlet that aligns with startup-level rewards", I think my statement of "millions of lines" is conservative.
Granted, not all of it was revenue-generating - much was experimental, exploratory, or just for fun. My overarching point was that I build software products for (great) living, as opposed to a marketer who stumbled into Claude Code and now evangelizes it as some huge unlock.
No, it’s not. At all. At the overwhelming majority of companies I’ve worked for or heard of, even 400-500 lines fully shipped in a week, slightly less than your figure here, would be top quartile of output - but further, it isn’t necessarily the point. Writing lines of code is a pretty small part of the job at companies with more than about 5-6 engineers on staff, past that it’s a lot more design and architecture and LEGO-brick-fitting - or just politicking and policying. Heck, I know folks who wish they could ship 400 lines of code a month, but are held back by the bureaucracies of their companies.
Now extrapolate. That’s maybe 50k a year assuming some PTO.
10 years would make 500k and you just cross a million at 20.
So that would have to be 20 years straight of that style of working and you’re still not into plural millions until 40 years.
If someone actually produced multiple millions of lines in 25 years, it would have to be a side effect of some extremely verbose language where trivial changes take up many lines (maybe Java).
i've been using llm-based tools like copilot and claude pro (though not cc with opus), and while they can be helpful – e.g. for doc lookups, repetitive stuff, or quick reminders – i rarely get value beyond that. i've honestly never had a model surface a bug or edge case i wouldn’t have spotted myself.
i've tried agent-style workflows in copilot and windsurf (on claude 3.5 and 4), and honestly, they often just get stuck or build themselves into a corner. they don’t seem to reason across structure or long-term architecture in any meaningful way. it might look helpful at first, but what comes out tends to be fragile and usually something i’d refactor immediately.
sure, the model writes fast – but that speed doesn't translate into actual productivity for me unless it’s something dead simple. and if i’m spending a lot of time generating boilerplate, i usually take that as a design smell, not a task i want to automate harder.
so i’m honestly wondering: is cc max really that much better? are those productivity claims based on something fundamentally different? or is it more about tool enthusiasm + selective wins?
Unless you're getting paid for your commute, you're just giving your employer free productivity. I would recommend doing literally anything else with that time. Read a book, maybe.
If you can't do your job in your 8 hours then you're either not good enough or the requirements are too much and the company should change processes and hire.
Right, I'm not saying anyone should actually be in the office 40 hours a week that sounds terrible. And even with all the RTO of the last couple years that doesn't seem to be expected many places.
Personally I use dev containers on a server and I have written some template containers for quickly setting up new containers that has claude code and some scripts for easily connecting to the right container etc. Makes it possible to work on mobile,but lots of room for improvement in the workflow still.
The project is just a web backend. I give Claude Code grunt work tasks. Things like "make X operation also return Y data" or "create Z new model + CRUD operations". Also asking it to implement well-known patterns like denouncing or caching for an existing operation works well.
My app builds and runs fine on Termux, so my CLAUDE.md says to always run unit tests after making changes. So I punch in a request, close my phone for a bit, then check back later and review the diff. Usually takes one or two follow-up asks to get right, but since it always builds and passes tests, I never get complete garbage back.
There are some tasks that I never give it. Most of that is just intuition. Anything I need to understand deeply or care about the implementation of I do myself. And the app was originally hand-built by me, which I think is important - I would not trust CC to design the entire thing from scratch. It's much easier to review changes when you understand the overall architecture deeply.
you can easily reach 50$ per day.
by force switching model to opus
/model opus
it will continue to use opus eventhough there is a warning about approaching limit.
i found opus is significantly more capable in coding than sonnet, especcially for the task that is poorly defined, thinking mode can fulfill alot of missing detail and you just need to edit a little before let it code.
