Projektor - an ai native jira / wiki tool self hosted on cloudflare. Currently considering how to extend it to cover more activities involved in software development https://github.com/TAJD/projektor

Iron volume, recently added a kettlebell complex generator which isn’t too bad if I say myself https://www.ironvolume.com/

I built my own very similarly targeted cloudflare hosted open source wiki and task management application https://github.com/TAJD/projektor

me three (at this point)

I've developed and deployed my own self hosted (on cloudflare / workerd) tool here https://github.com/TAJD/projektor and the key part was to be able to add it as a custom mcp server

the author referenced this paper which I thought was really interesting https://arxiv.org/abs/2307.03172 - "Lost in the Middle" - are there any other papers a bit like this that summarise lessons to do with applying llms

I'm not necessarily looking for the latest and greatest - more papers that those in the community have coalesced around providing nice ways of summarising problems or as a good example of a specific area.

A few things actually.

I’ve been working up reimplementations of Nikoli puzzles to host on my personal blog, the goal is to work through the maths and then explain how it’s relevant or helps create a fun games. Nikoli is the magazine that helped create sudoku and made it famous - they have a lot of other pen and paper puzzles that are really fun to play.

I’ve also got two other open source tools I’ve built to help me work daily:

https://tajd.github.io/cofferdam/ which is a rust based cli tool I wrote to parse the ast of codebase and enforce architectural preferences - eg don’t reimplement design system for components down to DRY

https://tajd.github.io/projektor/ Is an ai native project management tool where I can plan work on the train and then have my agents pick it up autonomously. Bootstrapped it quickly but then have been incrementally adding more functionality to it. It’s like self hosted jira on cloudflare but much faster and has more agentic development tools. I added tooling around being able to autonomously work through epics, manage quality and fix identified bugs - https://tajd.github.io/projektor/agents/playbooks/epic-goal/. The goal being deliver useful features and then to ultimately surface the number of bugs identified by agents and users so that I can see the health of the project and figure out if I need to write more code myself or improve feature descriptions to reduce discovered bugs. I don’t think this is new at all but it’s reliable and done well, I hope.

Also added a load more puzzle types to pokerchallenges.com - a website to train aspects of Texas hold’em - keen for feedback (from humans) so reach out and I’ll give you free access

Cloudflare is making it super easy to build projects and tools around it, so I thought I’d share my experiences of building my own tool for managing projects and fleets of AI agents.

Perhaps one interesting detail is about the distribution of the project. If I build and release the code as a binary it then becomes possible to have a config driven deployment of the tool, so it’s easy for people to redeploy it for their own data and requirements on top of cloudflare.

Obviously there is vendor lock in for now with cloudflare, but it’s possible to rewrite parts of the tool to run on open source workerd deployments and us other dbs and object storage.

Keen to hear what people think!

In real terms a product is the delivery of a service or other item of value. Specifically within the realm of software it is the promise of the delivery of software that will do what it says it will do, and then maintain or fix it where it falls short.

E.g. it's now about the delivery of a service rather than the challenge of building it (e.g. the moat) and then the delivery of service of it afterwards. Non functional requirements like data management and security, legal requirements and hosting are still valuable things that aren't front of mind for most developers creating their own tooling.

I think the art is about figuring out and defining what the underlying technical principles you want to leverage in order to provide a relevant software solution are. LLMs used naively might do that, but they probably won't. The work is still in figuring out what a useful domain representation of the problem is - and then you can unleash the llms for the boilerplate.

I completely agree - we've seen that tools with plugins are really scalable and then allow the better plugins to be folded back into the main tool where they become popular. I've taken that approach with my own dev tools that I've created.

And I've seen applications where the plugins are so damn complicated to make that it was easier to just make a separate tool to do that.

I think they were saying the "largest" output was the training, which I agree with. But it's also the case that they nudge forwards thinking in various areas through the outputs of the research process, whether that's papers, talks or the thesis itself.