Hi Stephen,

My apologies for the late reply. As you know, I've been experimenting with 
beads as well in TinkerPop for the last month or two, and I wanted to share 
some of my thoughts here as well.

Overall I agree that there is clear value in persisting key design decisions 
and motivations somewhere, which is currently lost as sessions are discarded 
and planning docs are never shared. In my experiments with it, beads seems to 
be quite a useful format for building plans, iterating on design, incremental 
implementations, and capturing outstanding questions or TODOs. I haven't done 
enough of an apples to apples comparison of alternatives to judge if beads 
truly has some sort of special sauce to it, but it has worked very well for my 
purposes.

I'm fully onboard with the rollout of an optional standardized beads workflow 
for TinkerPop. I think it's a great starting point to preserve more of this 
data, and it's a solid foundation to iterate and improve upon.

Thanks,
Cole

On 2026/04/17 15:44:47 Stephen Mallette wrote:
> I'd like to suggest optional use of Beads as a bit of a trial for
> contributors:
> 
> https://github.com/gastownhall/beads
> 
> I've been using it in conjunction with my coding agents and I like what it
> offers. I think it helps provide a bit more structure to how the agent's
> plans and then implements its plan. Initially, I just want to add something
> to AGENTS.md that lets the agent know that it should look for beads and use
> it if it is present. So in that sense it's optionally available. If
> contributors wish to use it, they can just install beads and it will be
> ready to go; otherwise, their agent will keep thinking in markdown.
> 
> Longer term, I think beads has the potential to move from optional to
> required depending on how well it is received here. i think it serves
> several purposes: (1) the sort of tactical one i mentioned above with agent
> planning/implementation (2) if beads were aggregated by contributors we
> would better track the activities of coding agents on their way to
> producing the code we accept from them. this information is currently left
> in markdown in our agent sessions and while some projects i think are
> asking that session be attached to pull requests, i don't think that's the
> best way to capture that information when there are tools like beads (3)
> offers some untold future benefits/insights as we effectively will build a
> a global beads graph - ah a graph! - of human/agent activities.
> 
> For now though - just suggesting a trial among contributors who want to
> participate and a relatively non-intrusive addition to AGENTS.md. If you're
> aware of other tools that might serve the purposes i've described above,
> i'd be curious to hear about them. Thanks!
> 

Reply via email to