lancekrogers

I was working with agents at scale, and the bottleneck moved from doing the work to reviewing, understanding, and tracking it. That was draining the time I wanted for thinking about the right thing to do.

Festival is a vibe engineering framework. It uses the filesystem and Git. You configure it to your workflows and processes. When it plans and executes, the work is done the way you want it done.

Everything lives in a workspace called a camp: the skills, planning documents, one-off scripts, and projects for one context. A festival is an ambitious goal inside that camp. Specs, smaller workflows, and ideas you intend to execute are tracked as work items.

You can work on many projects at once. Over time, camps function more like custom agents, where small seeds of your intentions trigger hundreds or thousands of individual actions.

Each camp has its own context and workflows, so switching camps changes how your agents work.

Fair point, COCOMO doesn’t tell you what a buyer will pay, it estimates what it would have cost a conventional team to build the same thing.

I’m using it strictly as a proxy for engineering effort, not market price. The takeaway isn’t “I made $7.8 M,” it’s “AI let one engineer produce work that historically required $7 – 8 M of salary-and-overhead to create.”

If you know a better public, third-party metric for comparing human vs. AI development effort, I’m all ears. The goal is to quantify the leverage, not to claim someone will write me an eight-figure check for the code tomorrow

I compared against recent employers codebases and the cocomo estimations where surprisingly close to the actual team size, timeline and cost. It's obviously not a perfect estimation, but it seems like the best model for comparing tools and communicating their impact.

My ROI so far is time savings and exponential increase in my personal production capacity. After I release Guild, I'm going to either productize it or focus on a few old startup ideas that were too time/capital intensive for me to bring to market without raising money, but now will only take me a few months to bring to market working part time

Why COCOMO?

I wanted a repeatable way to quantify the productivity jump I’m getting from Claude Code (and to compare it with other AI tools I’m testing). The line-counter scc prints COCOMO “organic” estimates by default; at first the dollar figures looked crazy, so I benchmarked them against a few past codebases where I know the head-count, timeline, and budget. They were surprisingly close, so I’m using COCOMO here as a rough yard-stick—not as a claim that LOC directly equals business value.

If nothing else, it gives engineers a concrete number to show when asking a boss to cover an AI subscription, or founders a way to justify “impossible” timelines to investors.

TLDR: Claude Code let me ship 219k high quality LOC in 7 weeks across eight projects, while juggling multiple distinct projects in parallel.

The problem is it will take years for the jobs to come back and a lot of people don't have years of liquidity.

If every developer is now 10x more productive, most businesses will be able to downsize until they start to be outcompeted by competitors who decided to build 10x better products rather than downsizing. The current norm is to keep the same productivity and shrink the workforce outside of small startups.

I've considered starting a handyman business if I truly can't find work in software again. I'd probably make more doing that than I've ever made from my actual job, the only difference would be I wouldn't have energy to focus on my startup ideas/side hustles.

I recently switched to lunarvim, it took a few weeks to port my neovim config over to lunarvim, but now it's perfect. My neovim config was a port from my .vimrc and was outdated, switching to lunarvim has been great.

I've used the jetbrains sweet of IDEs, vscode, atom and many other text editors and IDEs over my career but always come back to vim or neovim. After the initial hurdle of learning vim, everything is easier. My hands and shoulders hurt less and I can navigate code much more quickly. It also helps me stay more focused and actually remember what tools I have available and enables easy tweaking if the tools are broken or not 100% what I need.

I do project planning and time tracking in neovim now and have my notes synced across all devices using obsidian

It runs on mine ok as well, but the fingerprint scanner doesn't work

Yes