← narwal.one/Second Brain
SecondBrain
Ask the Brain
Index/Entityupdated Sat Aug 15 2026 08:00:00 GMT+0800 (Philippine Standard Time)

Sandeep Swadia

personcreatorthemitmonkai-literacyframeworks

Sandeep Swadia

Creator of the theMITmonk YouTube channel. Self-described background per his videos: "From homeless to MIT grad, then CEO / board member / investor in technology companies worth billions" and "advising AI companies worth billions." Newsletter at sandeepswadia.com. Distinct in this vault as the highest-framework-density single author — 6 videos so far, 11 named acronym frameworks.

Frameworks attributed to him in this vault

Framework Slot Purpose Source
TRAP Framework T-R-A-P How to learn anything durably How To Learn Anything So Fast (theMITmonk)
DRAG Framework D-R-A-G What to delegate to AI Dangerously Smart with AI (theMITmonk)
AIM Protocol A-I-M Quick prompting micro-pattern (Actor / Input / Mission) Dangerously Smart with AI (theMITmonk) (referenced as earlier; full source not yet in vault)
Intelligent Hill (Prompting Camps) 5-camp metaphor Prompting hierarchy: zero / one / few-shot → CoT → agents Dangerously Smart with AI (theMITmonk)
Intelligent Gym (no acronym) Deliberately frictioned AI use; progressive overload for the mind Dangerously Smart with AI (theMITmonk)
Intelligent Fool (no acronym) Beginner's-mind posture for using AI well Dangerously Smart with AI (theMITmonk)
PRIME Framework P-R-I-M-E Full-form prompting rubric (Purpose / Research / Interview / Mechanics / Examples) How To Use Claude Better Than 99% Of People (theMITmonk)
ARR Framework A-R-R Decide: agent or prompt? (Autonomous / Recurring / Reviewable) You're Not Behind (Yet) Learn AI Agents (theMITmonk)
GPS Check (for Agents) G-P-S Pre-flight before deploying an agent (Goal / Proof / Steps) You're Not Behind (Yet) Learn AI Agents (theMITmonk)
Narrow Agents (no acronym) The agents that win are obsessively narrow You're Not Behind (Yet) Learn AI Agents (theMITmonk)
DART Framework D-A-R-T Diagnose which system you're in (Deconstruct / Analyze / Recognize / Test) How To Think SO CLEARLY People Assume You're A Genius (theMITmonk)
Clear / Complicated / Complex / Chaotic (Cynefin Framework) 4-type taxonomy Match your response to the system's cause-and-effect type How To Think SO CLEARLY People Assume You're A Genius (theMITmonk)
ALTER Framework A-L-T-E-R Give AI five roles to build a personal university ("University in a Box"): Advisor / Librarian / Tutor / Editor / Roommate How To Become Dangerously Self-Educated With AI (theMITmonk)

Through-line across the videos

All four videos circle a single argument: AI compounds quality of thinking — good thinking gets faster results; bad thinking gets faster failures. The frameworks split into two clusters:

  • Human-side discipline — TRAP (learning), Intelligent Gym (deliberate friction), Intelligent Fool (beginner's mind), GPS (pre-flight clarity). What the user has to do to stay sharp.
  • Delegation and prompting — DRAG (what to outsource), AIM / PRIME (how to brief AI), Intelligent Hill (which technique to use), ARR (when to escalate from prompt to agent), Narrow Agents (where the commercial opportunity lives).

The pair "AI is a mirror; it amplifies the quality of your thinking back at you" (from the Agents video) is the load-bearing sentence underneath all of it.

Editorial pattern (stable across 4 sources)

  • Solo monologue, 13–18 min
  • 1–2 named acronym frameworks per video
  • Anchor anecdote (personal: homeless → MIT, hedge-fund call freeze, CMO meeting; or historical: Boyd, Nadella, Bjork)
  • Cognitive-science or military / business-history hook
  • Pithy single-line summaries that travel well as LinkedIn quotes
  • Sponsored episodes integrate the sponsor as a worked example of the framework (Remnote for Retain step; Higgsfield as MCP demo). Disclosed inline.
  • Closing on the human / philosophical note (Michelangelo, "you were the one you've been waiting for," "AI will make human life less robotic")

Why he matters to this vault

  • Highest cluster density of any single author here. Four sources, ~10 named frameworks, deeply cross-linkable. Many of the user-side concepts in this vault (Intelligent Gym, GPS Check) sit on a single Sandeep source — worth tracking whether they get triangulated by other authors before treating them as canonical.
  • The audience he's writing for overlaps the vault user's audience. Sandeep's videos are aimed at the ambitious knowledge-worker / senior-IC / operator / aspirant-founder — same band as the senior-IT-leader audience the user writes for. His frameworks are directly translatable into the user's brand-content workflow.
  • Frameworks are the unit of value. The videos themselves are quotable but the frameworks survive out of the video — they can be cited, taught, and applied independently. That's why this vault has more concept pages from Sandeep than from any other single author.

Synthesized in this vault

Sources

Update — 2026-06-13: systems-thinking / decision-making turn

How To Think SO CLEARLY People Assume You're A Genius (theMITmonk) adds a new cluster distinct from the prior AI-literacy frameworks: a decision-quality and diagnosis discipline under Systems Thinking (read hidden patterns from observable parts; parts → connections → patterns). Two new frameworks join the catalog above:

  • The DART Framework (Deconstruct / Analyze / Recognize / Test) — a diagnostic for placing a real, unlabeled problem into a system type.
  • The Clear / Complicated / Complex / Chaotic taxonomy, which closely tracks Dave Snowden's Cynefin Framework though Swadia teaches it without naming Cynefin.

He also surfaces the Cobra Effect (incentive traps) and delayed feedback loops as the recurring reasons systems confuse people. This is his first source that's squarely on the user's leadership/decision-making beat rather than AI upskilling — broadening his relevance to this vault beyond the human-side-of-AI cluster.

Update — 2026-08-15: AI-as-system-of-roles turn

How To Become Dangerously Self-Educated With AI (theMITmonk) introduces ALTER Framework (Advisor / Librarian / Tutor / Editor / Roommate) — the first Sandeep framework that's specifically about arranging AI as a multi-role system rather than a single-role tool. Where every earlier framework picked one slot (learning, delegation, prompting, agent design, systems diagnosis), ALTER structures the whole learning loop across five roles Sandeep argues must be run separately because their goals conflict if collapsed.

Second-order signals:

  • Sibling-video convergence. This video shares its title stem with How To Become Dangerously Self Educated with AI (Justin Sung) (ingested 2026-07-05). The two land within six weeks of each other and are directly complementary — Justin frames the why-skill (Higher-Order Thinking), Sandeep ships the how-stack (ALTER). Reads as a genre now, not a one-off.
  • Second MCP-in-the-wild demo. After Higgsfield in the Claude-99% episode, Zapier MCP is the worked example here — connecting the Advisor's plan to Google Calendar / Google Docs.
  • First NotebookLM deep-use. The Librarian role gets NotebookLM's clearest positioning in the vault yet, including its Studio-generated podcast + live-call-in feature.
  • First appearance of Two-Sigma Problem in the vault. Bloom's 1984 tutoring finding is the theoretical anchor Sandeep uses to justify the Tutor role — extending Sandeep's usual cognitive-science reference set (Ebbinghaus, Bjork, Bloom's Taxonomy) into education-effect-size literature.