← AI Catalyst C3All programsHomeSearch
AI Catalyst C3·Starter Kit·1:11:45

Onboarding Call: 'With AI, Not by AI' - the Six-Month Format, the No-Hand-Holding Rule, AI Burnout, and the Rented-Infrastructure Company

Outskill lead trainer Host - roll-call, philosophy, program logistics, Q&A · Paul Berg Participant - greeted in the roll-call; a chat question answered at 1:08

The short version

  1. The program: AI Catalyst, cohort 3, a six-month continuation for people who finished the Accelerator; explicitly beta - 'every cohort we change at least 20%'. Weekend main sessions of 2.5 to 3+ hours; a roughly three-week base camp of foundations, then weekday office hours; a Wednesday AI-updates session open to everyone; and invite-only sprints of two or three days on niche topics, announced with about a week's notice, kept 'outside curriculum'. LMS access, recordings and full transcripts for every session - 'feed the transcript to AI and make your own inferences' (0:29-0:41).
  2. The rules: nothing carries over from the Accelerator - 'all your tools, your perks, everything ended at the door'; frameworks and systems over tools; support, not hand-holding - worksheets are gone on purpose, this is 'a mental fitness gym'; no coding background needed; no promised 1:1 mentoring; a town hall for showing what you built (0:26-0:38).
  3. The philosophy: work with AI, not by AI - you stay the orchestrator ('Copilot, not autopilot'); guard the three capabilities AI can hollow out - critical judgement, creative synthesis, empathetic connection - and keep writing by hand (0:22-0:26, 0:36-0:38).
  4. High-functioning AI burnout: offload the easy and medium tasks and only the hard ones remain, so output rises while you tire - a Harvard-cited pattern; the countermeasure is deliberate non-AI time and one manual task a day (0:14-0:18).
  5. The rented-infrastructure company, Base44 as the case (sold for ~$400M with eight people, per the host's memory): ship a deploy every day, keep a supporter, a troller and a customer for real feedback, build only your differentiator and rent the rest (Stripe/Razorpay for payments, a platform for support); go hard niche and become the top 1% for a few people before broadening (0:42-0:48).

At a glance, three clicks deep

Skim here first: the closed row is the glance, open is the study card with the key points and timestamps, and the ↓ link drops to that concept's full write-up below.

01With AI, not by AI: you stay the orchestratorOrchestrate, don't delegate judgement;›

Orchestrate, don't delegate judgement; protect judgement, synthesis and empathy; write by hand sometimes.

'With AI, not by AI'; Copilot not autopilot (0:22-0:25)

Three capabilities to protect (0:36-0:37)

Pen-and-paper writing as practice (0:37-0:38)

Johari's window as the program frame (0:26-0:27)

↓ Full write-up of this concept

02High-functioning AI burnout: when only the hard tasks are leftOffloading easy work concentrates hard work;›

Offloading easy work concentrates hard work; schedule non-AI time and one manual task daily.

Easy/medium tasks gone -> only hard ones remain (0:15-0:16)

Higher output, more fatigue (0:16)

Non-AI time; one manual task a day (0:17-0:18)

↓ Full write-up of this concept

03How Catalyst runs: six months, weekend sessions, base camp, office hours, Wednesday updates, sprintsWeekend cores + base camp + weekday office hours + Wednesday updates + invite-only sprints;›

Weekend cores + base camp + weekday office hours + Wednesday updates + invite-only sprints; everything recorded and transcribed; 20% changes per cohort.

Six months; 2.5-3h weekend sessions (0:29-0:30)

Base camp -> office hours; Wednesday AI updates (0:32-0:34)

Sprints: 2-3 days, invite-only, ~7 days' notice, outside curriculum (0:33-0:35)

Transcripts for every session - 'make your own inferences' (0:40-0:41)

↓ Full write-up of this concept

04Support, not hand-holding: frameworks over tools, nothing carries over, no worksheetsTool-agnostic frameworks;›

Tool-agnostic frameworks; effort by design; WhatsApp + LMS + email; no 1:1 mentoring promised.

Nothing carries over from the Accelerator (0:26)

Frameworks over tools; niche tools go to sprints (0:27-0:28)

Worksheets removed; 'mental fitness gym' (0:35-0:36)

WhatsApp, LMS, email; no 1:1 (0:37-0:38, 1:06)

↓ Full write-up of this concept

05The rented-infrastructure company: Base44's three rules, and go hard nicheDeploy daily;›

Deploy daily; supporter + troller + customer; own only the differentiator; niche until you are the top 1%.

Base44: ~$400M, eight people (host's memory) (0:42-0:43)

One deploy per day (0:44)

Supporter, troller, customer (0:45)

Rent payments and support (0:45-0:46)

Hard niche first; pain from Reddit/Twitter (0:47-0:48)

↓ Full write-up of this concept

The concepts in full

01

With AI, not by AI: you stay the orchestrator

The one sentence the whole program is held to.

The human directs and pushes the AI; the AI does not make the decisions. The host borrows Microsoft's 'Copilot, not autopilot' framing and pairs it with the three capabilities to protect from atrophy - critical judgement, creative synthesis, empathetic connection - plus a concrete habit: hardcore pen-and-paper writing with no AI, to keep thinking clear. Johari's window (the four quadrants of knowing) is the frame for what the program is for: shrinking the unknown unknowns.

Why it matters

The behavioural bar every later session assumes.

02

High-functioning AI burnout: when only the hard tasks are left

You are producing more than ever and you are exhausted. The host says that is not a contradiction.

Citing a Harvard study, the host describes the pattern: AI absorbs the easy and medium tasks, so the human's day becomes wall-to-wall hard problems - output up, fatigue up. Countermeasures: deliberate non-AI time, and at least one manual task every day.

Why it matters

Self-care advice aimed precisely at the heavy AI user this cohort produces.

03

How Catalyst runs: six months, weekend sessions, base camp, office hours, Wednesday updates, sprints

The skeleton of the program, with the warning that it will change.

Six months; Saturday and Sunday sessions averaging 2.5 hours and stretching past three; a base camp of about three weeks on foundations, then weekday office hours for Q&A; a Wednesday AI-updates session for breadth; sprints - two or three days, invite-only, alumni-and-community, tagged beginner to advanced, run every two or three weeks with ~7 days' notice, deliberately outside curriculum. A two-week gap before base camp for LMS setup. Recordings and full transcripts for everything, with a promised session on building your own living knowledge base from them. Curriculum sneak peek: problem-solving with AI, first client and scoping, personal brand and niche, landing pages, prototypes and agents, weekly-report automation, prompting, advanced RAG, then full end-to-end products - 'expect a lot of sessions changing'.

Why it matters

The map of what the rest of this knowledge base is built from.

04

Support, not hand-holding: frameworks over tools, nothing carries over, no worksheets

'All your tools, your perks, everything ended at the door of the accelerator.'

Three rules. Catalyst teaches systems that work with any tool - 'systems that can make you use Copilot or any other tool' - and pushes niche-tool requests into sprints. Structured worksheets from earlier programs are removed on purpose: the struggle is the training - 'a mental fitness gym'. And no promises the program will not keep: no technical background required, no formal 1:1 mentoring, no fixed hackathon - but a town hall to show what you built. Support channels: the WhatsApp community for daily discussion, the LMS as the platform of record, email as the reliable route to the team (LinkedIn is not checked).

Why it matters

Sets expectations that explain the shape of every office hour that follows.

05

The rented-infrastructure company: Base44's three rules, and go hard niche

Eight people, one product, a ~$400M exit - if the host remembers the numbers right.

The Base44 story as told: a single founder to a small team, sold for roughly $400M with eight employees. Three rules drawn from it: ship at least one deploy a day; keep three kinds of feedback close - a supporter, a troller and a customer; and build only your core differentiator while renting everything else - payments through Stripe (or Razorpay in India), support through a platform. Then the niche rule: 'go hard niche, solve for only a few people, become the top 1% expert' before broadening, with pain points sourced from where people complain - Reddit, Twitter.

Why it matters

The one-person-company model the whole cohort is pointed at.

Tools referenced

ToolCoverageMomentContext
Copilotmentioned'Copilot not autopilot' framing
StripementionedRented payments
RazorpaymentionedRented payments, India
HermesmentionedNamed in the news rundown
OpenClawmentionedNamed in the news rundown

Action items

    Resources mentioned

    Resources
    • docPromised in chat
    • docAI-news rundown given as evidence the curriculum must keep changing

    Extraction notes

    This page was built from an auto-generated transcript, which garbles product and people's names. Those were corrected silently in everything above and logged here for transparency. The warnings flag claims that were true on the recording day but change fast.

    Transcript corrections applied

    The transcript saysThe trainer actually means
    CodecsCodex
    Pabhakar / Pabakarroll-call name, spelling unresolved
    Joanne (VC)author name as-heard

    True on recording day — verify before relying