summaryrefslogtreecommitdiffstats
path: root/2026/talks
diff options
context:
space:
mode:
authorSacha Chua <sacha@sachachua.com>2026-09-25 11:04:05 -0400
committerSacha Chua <sacha@sachachua.com>2026-09-25 11:04:05 -0400
commit5601a287a3c6587e6eb50dd1ceea5b9f3d7507c5 (patch)
tree4bb64ab7493ab95c905ae47cd758fe8ef51cdb5f /2026/talks
parent24fea6071941f20bdd2ea5bcb9eb51fb9f69c950 (diff)
downloademacsconf-wiki-5601a287a3c6587e6eb50dd1ceea5b9f3d7507c5.tar.xz
emacsconf-wiki-5601a287a3c6587e6eb50dd1ceea5b9f3d7507c5.zip
add team
Diffstat (limited to '')
-rw-r--r--2026/talks/team.md115
1 files changed, 115 insertions, 0 deletions
diff --git a/2026/talks/team.md b/2026/talks/team.md
new file mode 100644
index 00000000..ee3666cb
--- /dev/null
+++ b/2026/talks/team.md
@@ -0,0 +1,115 @@
+[[!meta title="Beyond the Chatbot: Team-Oriented AI in Emacs"]]
+[[!meta copyright="Copyright &copy; 2026 Mohsen BANAN"]]
+[[!inline pages="internal(2026/info/team-nav)" raw="yes"]]
+
+<!-- Initially generated with emacsconf-publish-talk-page and then left alone for manual editing -->
+<!-- You can manually edit this file to update the abstract, add links, etc. --->
+
+
+# Beyond the Chatbot: Team-Oriented AI in Emacs
+Mohsen BANAN (he/him) - Pronunciation: MO-HH-SS-EN, <https://github.com/mohsenBanan>
+
+[[!inline pages="internal(2026/info/team-before)" raw="yes"]]
+
+Subtitle: Durable, transferable AI context with the org-based bisos.aiActivity
+
+In these early days, none of us can claim to have figured out the right way
+to
+collaborate with our AI-Assistants (Claude, etc.). The common workflows are
+rudimentary: a back-and-forth chat that produces some result, which you
+then try
+to preserve in a CLAUDE.md; or "make a Plan, then build it." Integration of
+AI
+into our digital ecosystems and our usage environments remains primitive,
+individualistic, and experimental.
+
+Emacs is omni-moded. Over the past 40+ years people have added a mode for
+anything and everything, so layering AI-Assistants on top of all of those
+modes
+is the natural and instinctive model for Emacs. This omni-context,
+editor-centered, integrated usage environment now supports AI across all
+work
+contexts &mdash; software development (in any language), content creation and
+publishing (such as LaTeX), sophisticated email management (like Gnus), and
+task
+and time management (using org-mode) &mdash; consistently, naturally, and
+intuitively. The richness of this convivial aspect of Emacs is unique;
+those who
+think only in terms of the likes of VS-Code don't know what they are
+missing.
+
+Integration of AI into Emacs has multiple dimensions. This talk organizes
+them
+into three:
+
+1. Emacs packages, modes, and interfaces for talking to existing LLM-AIs &mdash;
+ gptel, aidermacs, claude-code.el, and many more.
+
+2. MCPs that expose Emacs' editing-oriented capabilities, and MCPs that
+ give AIs
+ full or partial access to the Lisp engine (and to Emacs applications) &mdash;
+ e.g. claude-code-ide.el and monet.
+
+3. AI-Workflows for establishing transferable, team-oriented long contexts
+ that survive across sessions, machines, and even a change of AI-Assistant.
+
+First I survey (1) and (2) &mdash; the fast-moving landscape of Emacs AI packages &mdash; and make the case that we do not yet have a framework for layering and modularizing these efforts.
+A great deal of duplicated and disjoint work is in progress.
+At this stage, that is not necessarily a bad thing.
+
+The emphasis of this presentation is on (3).
+Here the real problem is continuity across a discontinuity: Claude doesn't remember yesterday, and a colleague can read your code but not how you collaborated with your AI to produce it.
+I introduce an org-mode-based set of tools centered on bisos.aiActivity &mdash; a move beyond the chatbot toward AI that is convivial (shaped by its user, not the reverse), collaborative (many sessions and many colleagues toward shared goals), and cumulative (each participant profits from the others' AI interactions).
+
+aiActivity has two parts:
+
+1. bisos.aiActivity &mdash; a PyPI package (the thin execution engine),
+ installed with:
+
+ pipx install bisos.aiActivity
+
+ Sources: <https://github.com/bisos-pip/aiActivity>
+2. A set of templates &mdash; where the compounding value lives. The reference
+ templates tree is public and meant to be forked and adapted:
+
+ <https://github.com/bxexamples/aiActivityTemplates>
+
+The templates encode:
+
+- AI-Activity &mdash; the kind of work being done (e.g. bisos-pip, lcnt),
+ symlinked from a shared tree so refinements propagate to every
+
+project.
+
+- AI-WorkPlan &mdash; this specific effort, broken into Stages and TODOs.
+- AI-DevStatus &mdash; a working-state handoff to the next AI-session or
+
+colleague.
+
+- A set of Skills and a set of Commands.
+
+The model rests on org-mode, deliberately, not Markdown: real TODO state
+machines, scheduling, cross-referencing anchors, archiving, and dynamic
+blocks.
+Projects are decomposed with a hierarchical, tree-and-siblings model
+(initiate / initiateSub), so a base repo and its subprojects each carry
+their own
+plan while inheriting shared conventions. For Emacs package development, we
+can
+build directly on this.
+
+bisos.aiActivity originates in &mdash; and its default templates encode the
+conventions of &mdash; the Libre-Halaal ByStar/BISOS (ByStar Internet Services
+OS)
+digital ecosystem (<https://github.com/bisos>, <https://www.by-star.net>), whose
+Emacs-based usage environment is Blee (<https://github.com/bx-blee>). None of
+that is a prerequisite: the tool is org-mode based and the templates are
+open.
+Take them, adapt them &mdash; and if you try it, let me know what breaks.
+
+
+[[!inline pages="internal(2026/info/team-after)" raw="yes"]]
+
+[[!inline pages="internal(2026/info/team-nav)" raw="yes"]]
+
+