Frictionless local dev experience with permanent (or not) cloudflare tunnels github.com/cliftonc/cfld (I got tired of it being just one step too many!).
lastlight.dev/evals/introduc… ... now you can test your agentic coding workflows against historical 'golden' PR datasets (or public benchmarks); the only way to prove it's quality and make systematic improvements over time?
I started @fabriqaai with the goal of evolving it into a cloud-based long-horizon agent platform. However, the initial release was an Electron app that focused on the agent coordination layer first. For the past month, I've been running fabriqa on a cloud setup, allowing users to connect through the same chat interface from mobile devices or any browser. This setup enables agents to operate for hours on long, specification-driven tasks while conforming to defined tools, constraints, validation loops, and context systems.
I am currently working on transitioning fabriqa to run on ephemeral sandboxes, such as @e2b or other platforms, with the aim of creating agent teams that can operate with tens, if not hundreds, of agents running in parallel.
Had fun today exploring how i can test real world work in a harness with different models / workflows (early days, but I think this has potential): github.com/cliftonc/lastl…
@boyney123@flueai@mattzcarey I'm also working on migrating a full set of coding workflows from a project to flue; so far enjoying the experience if you were looking for a more complex end to end example: github.com/cliftonc/lastl…
Having so much fun with @flueai at the moment, writing custom agents for git workflows, from code > docs.
Found @mattzcarey project here github.com/mattzcarey/shi… for some nice example too 🙏
Ok, so just for fun I migrated my agentic SDLC harness (lastlight.dev) over to #flueframework from @FredKSchott ... really impressed, but trying to wrap my head around the primitives as I would have expected workflows to be durable? It works: github.com/cliftonc/lastl…
@eliakuratli If you need a richer way to visualise it I have an open source project that also runs fully on Cloudflare (try.drizzle-cube.dev) and even built a proof of concept for R2 Data Lake using it (try.icelight.dev) ... I am super bullish on the Cloudflare stack!
@ibocodes agree, i have built everything lately on cloudflare - but with one exception, d1 isn't that useful vs real postgres somewhere like Neon ... now if Cloudflare built a Neon competitor of serverless postgress i'd buy it in a heartbeat
hate to break it to you but most of your projects don't need
vps + postgres or supabase + vercel
just throw everything on cloudflare workers + d1
$5/month until you hit 10k users
stop overengineering
@marckohlbrugge@cursor_ai I think the answer is to put the work management where folks already are (github / linear), and not build yet-another-kanban-board; for example lastlight.dev/how-it-works/ (disclaimer: its my project)
I'm not sure a kanban board is the right way to manage your agents, but I do see think the current layout that every IDE uses (sidebar with sessions, main chat, canvas) lacks a global overview of everything you're doing. A long list of sessions is uninspired and unmanageable.
Also, right all IDEs treat agents as isolated instances rather than a piece of a bigger workflow.
Would love to see more design exploration in that
Who's working on this?
It's surprising to me that every developer is making their own "kanban to manage agents" but none of the major IDEs (@cursor_ai etc) have this UI built-in
Every open-source project should be engineering agent loops right now.
We've found success managing @warpdotdev with loops for:
- Issue triage
- Spec generation for larger features
- Code review
- Even letting agents self-improve their Skills on a cron
Want to add your own semantic layer to your app that makes your data fully accessible to agents / MCP without giving direct access to SQL? try.drizzle-cube.dev
For those deep in Claude Code every day and want to know why everyone else is annoyed that `claude -p` is now metered for subscription users: cliftonc.nl/blog/what-is-h…
@mattpocockuk For me (unfortunately), it just means moving more of my personal usage away from Claude - I built a PI wrapper that replaces my old usage of `claude -p` (in sandcastle-esque sandboxes) and it works great actually ... github.com/cliftonc/agent…