# SIG-P4B 2026-09-20 — Session Notes

Participants:
  - rafa (UTC+1)
  - Sachin

---

## 📖 The Reading
Not clearly identified in the discussion. The conversation centers on the participants' own workshop materials — an "AI CMM" (AI Capability Maturity Model) slide deck and a "kit" framework — rather than a shared external reading. The J.G. Ballard novels *Crash* and *Concrete Island* are referenced as conceptual comparisons but were not the session's reading.

## 🧭 Overview
This session was primarily a working/planning meeting between rafa and Sachin ahead of a workshop they are co-facilitating the next day. It opened with a conceptual discussion refining the definition of "kits" and the role of protocolization in the AI environment, then shifted into practical preparation: facilitation roles, agenda timing, attendee logistics, and the state of the shared GitHub/Discord setup and "context tank." Much of the second half is operational detail about registrations, calendar invites, and exercise sequencing.

## 💡 Key Points & Themes
- **Sachin — refining the "kit" definition via the car analogy:** He revisited the comparison between AI kit-building and early car tinkering. He argued that early cars existed in a "very low protocolization environment" (no roads, highways, traffic laws, or authorities), whereas AI kit work happens in a "much more heavily prototyped/protocolized world."
- **Sachin — a better car analogy:** He suggested Ballard's *Crash* and *Concrete Island* as better comparisons, since those depict figuring something out within an environment that is already heavily built up, rather than starting from scratch.
- **Sachin — evidence for protocolization framing:** He grounded the claim in a Protocol Institute graph showing rising usage of the word "protocol" after the Second World War.
- **Sachin — distinguishing protocolization from skill environments:** Early car tinkerers already had machine-tinkering skills; that reflects an environment where relevant skills existed, which he distinguished from true protocolization. By contrast, AI sits atop many existing protocols (e.g., Git/GitHub, various frameworks).
- **rafa — is protocolization relative/perceptual?** He pushed back by asking whether we only see our environment as "highly protocolized" from our vantage point — would a farmer of that era have felt they were in a protocolized environment? He speculated we might actually be in a comparatively *non*-protocolized environment we can't yet perceive, and that in fifty years everything may look "hyperprotocolized" by comparison (analogous to a farmer being unable to conceptualize farm automation).
- **rafa — AI as "oozy":** He noted AI is special because it is "oozy" and itself protocolizes environments in some ways, suggesting the dynamics differ from the car case.
- **Sachin — smooth vs. striated:** He connected this to an oscillation between "smooth and striated" — AI can make already-striated fields "smoother and oozier," and vice versa.
- **Maturity levels:** Sachin argued that maturity levels make sense precisely in an environment that already has many protocols, where you need to manage the introduction of new ones — tying the theoretical discussion back to the AI CMM.

## 🔀 Questions & Disagreements
- **Is protocolization objective or perceptual?** The main friendly disagreement: Sachin leaned on the Protocol Institute graph as a rubric for increasing protocolization over time, while rafa questioned whether "protocolized" is a relative judgment that we can't fully assess from inside our own moment.
- **Whether to include failure modes / success patterns (AI CMM slide 25) in tomorrow's workshop:** They were unsure whether this content belongs in workshop one or is better saved for workshop two. They agreed it may be "too much" for the first session but decided to keep the slides on hand and pull them up if there's a right moment.
- **How to handle a large turnout:** Open question about what to do if all ~30 registrants attend — possibly group breakouts, though doing breakouts on Discord was flagged as potentially difficult. They tentatively resolved to keep the main session in the Kafka room with flexible breakout rooms, and to add a Google Meet as a fallback.
- **Exercise timing:** Some back-and-forth on how long the inventory exercise should take (~ten minutes) versus other exercises; left somewhat flexible.

## 🔗 References Mentioned
- J.G. Ballard novels — *Crash* and *Concrete Island* (as analogies for figuring things out in an already-built-up environment).
- The Protocol Institute — specifically a graph showing rising use of the word "protocol" after WWII.
- "AI CMM" / AICMM — their AI Capability Maturity Model slide deck (slide 25 referenced for failure modes / success patterns).
- Git / GitHub — cited as examples of existing protocols underlying AI work; also their repo for workshop materials.
- Discord — hosting the workshop (Kafka room; ~nine breakout rooms available).
- Google Forms — registration sign-up; Google Meet/calendar invites for logistics.
- "Meat proxy" (transcribed "Meet proxy") — a derogatory term rafa saw on Twitter for someone who doesn't think for themselves and just uses ChatGPT; they also joked about "MPC/NPC"-style memes.
- "Context tank" — a shared resource/repository the group contributes to.
- Concepts: "smooth vs. striated," "kits," "factories," "maturity levels."
- Third parties named: Michael Somby (University of Exeter, last registrant); Patrick / "Pat Water" and Wip Abramson (participants who pushed inventories); "Will" (referenced re: adding more kits — likely the same as Wip Abramson).

## ✅ Action Items & Next Time
- **rafa:** Email the last registrant (Michael Somby); send calendar invites (with Google Meet fallback) to all registered emails; upload a snapshot of the current slides to the GitHub repo before attendees start "messing around" with it.
- **Both:** Confirmed calendar/timing (workshop at ~10:30; rafa has a robotics workshop just beforehand and will try to leave it early). Verified the Kafka Discord room link works.
- **Facilitation plan for session one:** Sachin does an intro on the first slide; rafa covers the Protocol Institute and logistics; Sachin leads the "kit theory"; then hands back to rafa for the exercises. Rule of thumb — theory questions go to Sachin, logistics questions go to rafa. Session two expected to be more of a back-and-forth.
- **Exercises:** Inventory exercise (~10 min) is the main first activity, aiming to have participants find more than one kit (ideally ~5); followed by participants reviewing each other's kits, then setting up the "factory." Push Wip/Will to add more interconnected kits beyond the two he listed. Keep the agenda flexible and let the audience steer direction.
- **Idea for a future version:** rafa floated recording the full workshop so an agent could auto-improve the next version; Sachin suggested a RAG agent could run through the workshop "exhaust" for next time.
- **No new content** to be added to the context tank ("all we have is good").
- Next concrete event: the workshop itself, the next day.

## ⭐ Memorable Quotes
- **rafa:** "Maybe we might be in a very nonprotocolized environment and we just can't see it... in fifty years everything is gonna be, like, hyperprotocolized, and we just, like, can't even conceptualize it."
- **Sachin:** "AI makes some fields that were already striated much more smooth and oozy, and [the] other way around as well."
- **rafa:** "I think it'll be relatively quick... because your AI is doing it."