Argument
GPT-5’s architecture — a distributed routing system that allocates cognitive resources across specialized processors — is structurally analogous to how neurodivergent (specifically ADHD-autism) brains process information. People who spent decades learning to translate between non-linear, parallel cognitive styles and neurotypical communication norms have inadvertently trained themselves in the exact skills that unlock GPT-5’s advanced capabilities. The “value gap” between novice and expert AI users mirrors the gap between neurotypical and neurodivergent processing — and for once, the neurodivergent approach is the optimal one.
Structure
Four sections following the newsletter’s Spark / Pattern / Protocol / Debug format:
- The Spark — Framing: while others complained GPT-5 was “dumbed down,” the author found it remarkably capable. The difference was thirty years of debugging a “distributed intelligence system” that everyone else called being “weird.”
- The Pattern — OpenAI’s background router system (allocating tasks across base, thinking, and pro processors) is described as “the author’s internal monologue made manifest.” The “value gap” between expert and novice users is the same gap the author has lived in their whole life.
- The Protocol — Four specific GPT-5 techniques derived from neurodivergent communication experience: (1) trigger words for resource allocation (“Think deeply about this”), (2) agentic eagerness control (autonomy levels for creative vs. technical work), (3) reasoning effort specification (“Apply maximum reasoning effort”), (4) XML-style structured prompts. Bonus: “Octopus Mode” — a full ultra-deep-thinking prompt that forces multi-angle verification and self-critique.
- My Debug — Reflection: mastering GPT-5 was about encountering a system finally sophisticated enough to understand the author’s native communication style.
Key Examples
- GPT-5’s “background router” — allocates requests across three specialized processors (base, thinking, pro) based on complexity signals.
- “Octopus Mode” prompt — extended prompt engineering template that forces GPT-5 to break tasks into subtasks, challenge its own assumptions, triple-verify, and explicitly document edge cases. Full prompt text included in the piece.
- XML-structured prompts —
<task>,<context>,<output_format>,<reasoning_level>tags for eliminating ambiguity. - The “value gap” in AI — the same researcher-documented gap between expert and casual AI users that the author has lived as the neurodivergent/neurotypical gap.
Connections
- You Don’t Need Better AI. You Need Better Add-Ons — the prior piece on general prompt enhancers; this piece follows up with GPT-5-specific techniques
- Octopus Mode — wiki page likely exists for this concept given the broader wiki content
- GPT-5 — subject of the piece
- OpenAI — builder of GPT-5 and the routing architecture described
What It Leaves Open
- Whether GPT-5’s routing system actually works as described or whether the “distributed intelligence” framing is the author’s metaphor rather than OpenAI’s documented architecture.
- Whether the neurodivergent communication advantage will persist as models improve, or whether future AI becomes accessible to all communication styles equally.
- The “value gap” claim: the piece asserts it but doesn’t cite specific research on the expert/novice gap in GPT-5 specifically.
- Ethical questions about AI systems designed to require expert prompting: who gets left behind when advanced AI capability requires specialized knowledge to access?
Newsletter Context
Second in an informal two-piece sequence on prompt engineering (preceded by “You Don’t Need Better AI. You Need Better Add-Ons.”). More personal and opinionated than the prior piece — less a practical guide than a reframing of neurodivergent identity as technical advantage. The “Octopus Mode” prompt is the most actionable piece of content, included verbatim. Sits on the AI/technology beat but is also a piece about neurodivergence, self-understanding, and institutional recognition failure.