Why even have a human-oriented interface and not something more graph like?
Why not move the decision making closer to where humans are actually working and making those decisions (e.g. Slack, Teams)?
I think the Agensh paper[0] has the most interesting take: it gives agents a Mattermost instance to coordinate work against. There's no reinventing of software tickets for the AI era; I don't think the AI era needs tickets which are text copy-pasted from where the actual decision was made. Why take decisions made in a Slack or over email or in the transcript of a call and then do one more transcription into a ticket (except that a process calls for it today)?
Claude Meetings feels directionally correct: move the agent into the human workspace.
Interesting feedback. I liked the post. I see what is being advocated as a workable near term "hack" to mitigate some of the problems. What you suggest may in fact be a viable longer term option or at least part of a longer term approach.
Key distinction I believe: Inference alone can't get us there because there is a difference between Agents [properly] inferencing decisions and humans actually endorsing / authorizing the presumed decisions. So, what the post gets right is that humans have to do something explicit to articulate key decisions guiding the work and therefore we need some verifiable and traceable artifact back to the decisions, who made those decisions, and also ideally why.
Beyond that, this structure should really exist outside the silo of the agents doing the work in some kind of standardized form accessible to all parties involved in the work (people and agents)...which I think the post also got right.
And I agree this goes well beyond agents. This is a step towards capturing decisions in a shared form that's accessible to both humans and agents, while making those decisions easy to discover and retrieve when they become relevant.
Agents just make the need more urgent, given the speed and scale at which they operate.
> So, what the post gets right is that humans have to do something explicit to articulate key decisions guiding the work and therefore we need some verifiable and traceable artifact back to the decisions, who made those decisions, and also ideally why.
That's what the Slack/Teams thread is for; the agent doesn't have to be a "read-only" participant; it can and should be an active participant. Asking questions, determining if it's final, and doing the dirty work of record keeping.
That's why the PR feels more appropriate. Decision is made. PR is opened. PR has summary and link to the thread(s) where the real discussion occurred and real decisions were made, tradeoffs discussed, etc. A ticket is just another text collaboration surface and one that is less dynamic and less fluid. It becomes just another copy-paste proxy for the real stream of decisions.
Good post. Fully agree with the premise of needing to formally document decisions and also like the clever way of structuring tickets in Linear to help address this.
I have been thinking about this for several years as the problems here are not specific to agents. It happens anytime people are coordinating to get things done...it's just that humans are better at, and used to, working around this via various ad hoc approaches.
In fact, the broader concept this related to - referred to as "Articulation Work" dates back to the 1980's in the area of Computer Supported Coordination Work (CSCW) research.
There is a seminal paper from 1992 by the researchers Kjeld Schmidt and Liam Bannon titled "Taking CSCW Seriously: Supporting Articulation Work." and I posted about it from different angle recently on our own company blog.
This doesn't seem right to me.
Why not just PRs instead of tickets?
Why even have a human-oriented interface and not something more graph like?
Why not move the decision making closer to where humans are actually working and making those decisions (e.g. Slack, Teams)?
I think the Agensh paper[0] has the most interesting take: it gives agents a Mattermost instance to coordinate work against. There's no reinventing of software tickets for the AI era; I don't think the AI era needs tickets which are text copy-pasted from where the actual decision was made. Why take decisions made in a Slack or over email or in the transcript of a call and then do one more transcription into a ticket (except that a process calls for it today)?
Claude Meetings feels directionally correct: move the agent into the human workspace.
[0] https://arxiv.org/abs/2609.26781
Interesting feedback. I liked the post. I see what is being advocated as a workable near term "hack" to mitigate some of the problems. What you suggest may in fact be a viable longer term option or at least part of a longer term approach.
Key distinction I believe: Inference alone can't get us there because there is a difference between Agents [properly] inferencing decisions and humans actually endorsing / authorizing the presumed decisions. So, what the post gets right is that humans have to do something explicit to articulate key decisions guiding the work and therefore we need some verifiable and traceable artifact back to the decisions, who made those decisions, and also ideally why.
Beyond that, this structure should really exist outside the silo of the agents doing the work in some kind of standardized form accessible to all parties involved in the work (people and agents)...which I think the post also got right.
And I agree this goes well beyond agents. This is a step towards capturing decisions in a shared form that's accessible to both humans and agents, while making those decisions easy to discover and retrieve when they become relevant.
Agents just make the need more urgent, given the speed and scale at which they operate.
That's why the PR feels more appropriate. Decision is made. PR is opened. PR has summary and link to the thread(s) where the real discussion occurred and real decisions were made, tradeoffs discussed, etc. A ticket is just another text collaboration surface and one that is less dynamic and less fluid. It becomes just another copy-paste proxy for the real stream of decisions.
[dead]
[flagged]
Good post. Fully agree with the premise of needing to formally document decisions and also like the clever way of structuring tickets in Linear to help address this.
I have been thinking about this for several years as the problems here are not specific to agents. It happens anytime people are coordinating to get things done...it's just that humans are better at, and used to, working around this via various ad hoc approaches.
In fact, the broader concept this related to - referred to as "Articulation Work" dates back to the 1980's in the area of Computer Supported Coordination Work (CSCW) research.
There is a seminal paper from 1992 by the researchers Kjeld Schmidt and Liam Bannon titled "Taking CSCW Seriously: Supporting Articulation Work." and I posted about it from different angle recently on our own company blog.
[flagged]