Ask HN: What are you working on? (January 2026)

Holidays nuked all the hot-cached context in my head. I spent a few days just spinning wheels until it repopulated. But the basic idea works now!

Much testing and benchmarking work remains to make sure it's not going to lose data, and that it won't denial-of-service itself (because object-map -> facts fan-out is big).

Also a second giant blog post is due (following the one discussed above). Lots of notes have accumulated.

It will be fun even if the concept ultimately crashes and burns to the ground :)

In which case, there's always datomic and xtdb :D

  (def repl-facts
    [{:alt.site.evalapply/meta
      {:description "Root namespace for a named site."}}
     {:alt.site.evalapply/features
      {:paid #{:alt.site-feature/feat-1
               :alt.site-feature/feat-2
               :alt.site-feature/feat-3}
       :trial #{:alt.site-feature/feat-4}
       :complimentary #{}
       :available #{}}}
     {:alt.site.evalapply/users
      {:authorised #{:alt.user.evalapply/user-1
                     :alt.user.evalapply/user-2
                     :alt.user.evalapply/user-3}
       :unauthorized #{:alt.root.*}}
      :alt.user.evalapply/user-1 {:name {:first "Wiley"
                                         :last "Coyote"}
                                  :roles #{:alt.role.evalapply/owner}}
      :alt.role.evalapply/owner {:rw #{:alt.user.evalapply/*}
                                 :ro #{}}}])

  (assert-facts! (#'user/system-state)
                 ::app
                 :alt.root/user-1
                 (uuid/v7)
                 repl-facts)

  (redact-facts! (#'user/system-state)
                 ::app
                 :alt.root/user-1
                 (uuid/v7)
                 (subvec repl-facts 2))

  (read-now! (#'user/system-state)
             ::app)

As a hardcore AI chat user, I'm often frustrated with the single-agent workflow, where a single context window is used for even very long conversations. If I want to change the topic, open a thread, or go on a tangent, I often end up compromising the main thread and I'm forced to copy context over if I want to dive into something.

To solve this, I'm working on a collaborative AI agent orchestrator that models the solution as a group chat with humans and AI agents, including an agent orchestrator.

You can spawn participating agents with the orchestrator who will decisively route messages to the existing agents, or spawn new agents if needed. Also, you can open agent details and send messages directly to existing agents, similar to threads in slack.

So far, I have MCP integrations working with Linear and GitHub, but plan to add many more.

I've been working on this just over 2 weeks, making heavy use of 4+ concurrent Claude Code agents. This would have been impossible otherwise.

If you're interested, feel free to DM on X.

https://x.com/jon_ator/status/2010370649147998459?s=20

Obviously, it all comes with its own sets of issues, but I am working through those as they come. But it is still a slow move solo.