Cohort 01 · 40 seats · Live on Zoom

Get dates first →
Vol. 01 · The Five Day Founder Method

Build an AI-first
business with Claude.
In five days.

From vague idea to live product, brand, payment flow, and AI operating system — on the newest stack: Fable 5, Claude Code, Claude Design, Cowork, Goals, Loops, Workflows, Skills, and MCP. No coding experience required.

Most founders use Claude to answer questions. You’ll use Claude to research, design, build, deploy, monitor, and operate.

Forfounders starting from zero.
Format
Five evenings · live on Zoom
Session
60 min live + ~2 h guided build
Cohort
01 · 40 seats · from AUD $199
Dates
Announced to the list first
The syllabus stack · rewritten July 2026
  • Fable 5
  • Opus 4.8 + fast mode
  • Sonnet 5
  • Claude Code
  • Claude Design
  • Goals
  • Loops
  • Dynamic workflows
  • Ultra mode
  • /code-review ultra
  • Cowork
  • Skills
  • MCP orchestration
  • Managed Agents
  • Plugin marketplace
  • Persistent memory
A founder's quiet workspace at golden hour - laptop, notebook, coral ribbon
Pl. 01 · The table where it gets built

§ 02 · The Thesis

Most people use Claude like a chatbot.

You’ll use Claude like a company.

In the last few months, Claude has changed from a conversation tool into a working stack: Fable 5 for the hardest reasoning and week-long work, Opus 4.8 with fast mode for everyday building, Sonnet 5 for agent fleets, Claude Design for prototypes and visual systems, Claude Code for building — now with goals, loops, and dynamic workflows that keep it working until the job is verifiably done — Cowork for knowledge work, Skills for repeatable workflows, and MCP for connecting real tools.

The problem is not that Claude is underpowered. The problem is that most founders are only using 5% of it.

Most founders haven’t caught up. This course is the catch-up.

The Release Log · Why this cohort exists now

The stack moved again while you were busy building.

Most people have seen the announcements. Almost nobody has combined them into a working business system. This is the actual release log of the last ninety days — keeping up with it is a full-time job you didn’t apply for.

Compiled 11 Jul 2026 · anthropic.com/news · openai.com

  1. 28 May

    Claude Opus 4.8

    New flagship Opus - four times less likely to let a flaw pass unremarked.

  2. May

    Dynamic workflows

    Claude Code writes its own multi-agent orchestration scripts - migrations, audits, bug hunts.

  3. 09 Jun

    Claude Fable 5 & Mythos 5

    A new Mythos-class tier above Opus. The most capable model ever made generally available.

  4. 30 Jun

    Claude Sonnet 5

    Near-Opus agentic coding at a fraction of the price - the workhorse for agent fleets.

  5. 09 JulElsewhere

    OpenAI GPT-5.6 - Sol · Terra · Luna

    The other frontier lab answers, in Codex and Copilot the same week.

  6. H1 '26

    Goals · Loops · Ultra mode · Ultra review

    Claude Code learned to keep working - self-pacing, self-checking - until the job verifiably holds.

This course turns the feature avalanche into a founder workflow.

§ 03 · The Deliverables

What you’ll ship by Friday.

Not a toy demo. Not a prototype you’ll abandon. A product that loads, signs users in, stores real data, has a Stripe checkout flow, and looks like a real company built it.

A sketched paper mockup of a product landing page with a coral wax seal
Pl. 02 · From sketch to shipped
  1. 01A validated business idea with a real market
  2. 02A live MVP on a public URL
  3. 03Auth, database, and Stripe checkout flow
  4. 04A complete brand system and launch assets
  5. 053 custom Claude Skills for your business
  6. 063 scheduled Claude workflows for growth, maintenance, and operations
  7. 07A personal AI operating system for running the business

The Friday Ship Guarantee

Attend all five live sessions, complete the daily build checkpoints, use the approved starter stack, and ask for help before the final session if you’re blocked.

If you still don’t have a deployed working MVP by Friday, you get a full refund or a free retake in the next cohort.

Tool truth

The newest stack, without pretending every feature is magic.

Judgment is the curriculum underneath the curriculum. You’ll learn:

  • 01when to use Claude Design vs Claude Code
  • 02when to use Cowork vs Claude Code
  • 03when a /loop beats a Routine - and when a goal should refuse to let the agent stop
  • 04when ultra mode is worth the tokens, and when it's theatre
  • 05when to trust /code-review ultra, and when to read the diff yourself
  • 06when MCP is useful and when it is dangerous
  • 07when to keep a human approval step
  • 08when a simple spreadsheet beats a complicated agent
  • 09when not to connect production systems
§ 04 · The Curriculum

Five days. Five deliberate sessions.

Each day is a 60-minute live session on Zoom, then roughly two hours of guided homework you do on your own. Direct and outcome-led, no filler.

Five cream index cards numbered 01 to 05 arranged in a row
Pl. 03 · The week, indexed
01Day 01 · 60 min live

Ideation & Market Intelligence

From I have an idea to I have a validated wedge.”

Turn a vague itch into a precise wedge - a thing you can sell, to a person who actually has the problem, in a market that can pay. We use Claude the way an analyst uses an analyst team. You'll build a "Market Intelligence Room" in Claude: project instructions, uploaded notes, competitor links, pricing screenshots, customer-review exports, and a persistent market map.

Core tools  Research · Projects with RAG · Cowork Projects · Artifacts (persistent) · Custom visuals in chat · Research citations · Market monitor setup · File Creation

02Day 02 · 60 min live

Build & Deploy Your MVP

From I have a validated idea to My product is live on the internet.”

From blank folder to deployed URL. We use Claude Design to shape the visual language, Claude Code to write it in Plan Mode, and shadcn + Vercel MCP to ship. Then you'll set a goal - "a working MVP is deployed and the homepage loads clean" - so Claude can't call it done until it verifiably is, and a /loop deployment babysitter that checks logs and the live page on its own schedule.

Core tools  Claude Design · Claude Code · CLAUDE.md · Plan Mode · Fast mode · shadcn/ui · Vercel MCP · Claude in Chrome · /goal ship conditions · /loop deployment babysitter

03Day 03 · 60 min live

Database, Auth & Payments

From My product is live to My product is a business that can take money.”

The difference between a demo and a company. Supabase for data, Clerk for identity, Stripe for revenue - wired together by Claude with sub-agents and hooks, then checked by /code-review ultra before anything touches live mode. We use Stripe test mode during the cohort and give you the live-mode checklist for launch week.

Core tools  Supabase MCP · Clerk MCP · Stripe MCP · Vercel env vars · Sub-agents · Hooks · /code-review ultra · RLS review · Secrets management · Prompt-injection checklist

04Day 04 · 60 min live

Brand & Discoverability

From I have a working product to I have a company that looks credible.”

The thing people forget. Visual identity, a website that closes, launch assets that look investor-ready, and the discoverability system that keeps compounding after launch. All through one shared Claude context.

Core tools  Claude Design · Design systems · Claude for Word · Claude for Excel · Claude for PowerPoint · Artifacts · Custom visuals · DataForSEO MCP · Google Search Console MCP (optional) · Image-generation MCP (optional)

05Day 05 · 60 min live

Your Personal AI Operating System

From I do everything manually to Claude runs the weekly machine.”

The final, quiet superpower. You'll leave with custom Skills, MCP connections, scheduled Cowork tasks, browser workflows, Claude Code Routines - and goals, loops, and dynamic workflows that keep your business moving after the cohort ends, pacing themselves and refusing to stop until the work holds.

Core tools  Custom Skills · Skill Creator · Claude Code Routines · Goals & loops · Dynamic workflows · Cowork scheduled tasks · Chrome scheduled workflows · Dispatch · Computer use · MCP orchestration · Managed Agents (optional advanced)

§ 05 · The Fit

+

This is for you if

  • You have an idea and want to ship it this week
  • You're a consultant, operator, or domain expert without code
  • You're a founder who wants to go 10× faster
  • You're done with tutorials and ready to build

Honest about fit

This isn’t for you if

  • You want passive inspiration without shipping
  • You're looking for a traditional coding bootcamp
  • You don't have 5 hours live + 10 hours homework this week

No hard feelings — join the list below and catch the next one when the week is yours.

§ 06 · In Practice

Not a lecture. A workshop.

You’ll be in Claude Design, Claude Code, Claude for Excel, Cowork, and the MCP ecosystem every day - not reading about them. A few moments from the curriculum, shown in context.

A laptop showing a design canvas with a Claude assistant sidebar suggesting swatches and typography
Fig. 01 · Claude Design

Claude Design

A brand system, by Wednesday.

Moodboard, palette, typography, component library, hero mockup. Claude Design builds the visual language while you keep writing copy - one shared context, zero Figma tax.

Claude Design · Artifacts · Brand brief template

A laptop showing the Claude Code terminal running an agentic coding plan with step 3 active
Fig. 02 · Claude Code

Claude Code

Your new staff engineer.

Plan Mode first, write second. Claude Code reads your CLAUDE.md, plans the change, shows you the diff, and ships it - with hooks guarding anything destructive. Set a goal and it won't call the job done until the ship condition verifiably holds. You stay in charge; it does the typing.

Claude Code · Plan Mode · CLAUDE.md · Goals · Hooks · Sub-agents

A laptop on a quiet evening desk running a scheduled Claude Code routine with timestamped log entries
Fig. 03 · Claude Code Routines & Loops

Claude Code Routines & Loops

Your product's night shift.

Set a routine once: review new issues, check deploy logs, inspect failed builds, open a draft PR, or summarise what changed this week. Or hand it a /loop with no interval and let it pace itself - polling the deploy while it's fragile, backing off when it's calm.

Claude Code · Routines · /loop dynamic mode · GitHub triggers · Scheduled reviews

A laptop showing a spreadsheet with the Claude assistant sidebar generating LTV formulas
Fig. 04 · Claude for Excel

Claude for Excel

Your numbers, narrated.

Drop a messy CSV in front of Claude. Get a clean unit-economics model, a formula you understand, and a plain-English explanation of what each row means. Then ship it.

Claude for Excel · Artifacts · Inline Visualizer

A Claude Cowork dashboard with parallel agents scheduled across the day
Fig. 05 · Claude Cowork

Claude Cowork

Agents that run on schedule.

Four agents, scheduled and accountable. Your inbox gets triaged, your metrics get reviewed, your leads get enriched, your content gets drafted - every morning, before you're at the desk. Some run in the cloud. Some run from your desktop. You'll learn the difference - and when each is safe.

Cowork · Scheduled tasks · Dispatch · Sub-agents

A hand holding a phone with a Claude chat while a laptop in the background shows dispatch active
Fig. 06 · Dispatch

Dispatch

Delegate from your phone.

Message Claude while you're away from your desk. It can use your desktop files, configured connectors, browser, and apps - then send back the finished memo, spreadsheet, pull request, or briefing.

Dispatch · Cowork · Computer use · Mobile delegation · Local files

A hand-drawn MCP paper diagram showing Claude at the centre wired to six product tools
Fig. 07 · Model Context Protocol

Model Context Protocol

The tool layer for your AI company.

MCP is how Claude stops waiting for copy-paste and starts working with your real stack: database, payments, deployment, analytics, documents, and internal tools. By Friday, you'll know how to connect MCP servers, decide what should stay read-only, and avoid the dangerous setups that expose production data.

Supabase MCP · Stripe MCP · Vercel MCP · DataForSEO MCP

The Four Automation Layers

By Friday, you’ll know when to use each layer.

The course teaches architecture, not just tools. Some automation runs in the cloud. Some runs from your desktop. Some runs in the browser. Some paces itself and refuses to quit early. You’ll learn the difference — and when each is safe.

01

Cowork scheduled tasks

Best for knowledge work on your desktop.

  • Reports
  • Inbox summaries
  • File organisation
  • Lead research
  • Document preparation
02

Claude Code Routines

Best for product and engineering workflows.

  • PR reviews
  • Bug triage
  • Deployment checks
  • Dependency audits
  • Docs drift
  • Issue triage
03

Chrome scheduled workflows

Best for repetitive browser work.

  • Checking dashboards
  • Pulling search data
  • Reviewing competitor pages
  • Submitting routine forms
04

Goals & loops

Best for work that must not stop until it's done.

  • Deployment babysitting
  • Self-pacing monitors
  • Overnight bug hunts
  • “Done means done” ship conditions
The Friday OS

By the final session, you’ll install the operating rhythm behind your startup.

The goal is not to automate everything.

The goal is to stop doing the repetitive work that keeps founders from building.

Daily

  • Lead research
  • Inbox triage
  • Product-health check
  • Founder briefing

Weekly

  • Competitor monitor
  • SEO review
  • Content repurposing
  • Customer-feedback synthesis

Monthly

  • Pricing review
  • Landing-page audit
  • Roadmap review
  • Investor / customer update
Built live on Day 05 · yours to keep
§ 07 · The Rate Card

Three ways to join. Same method.

Pick your tier, pay once, turn up on Day 1. Full refund if you don’t have a deployed product by Friday.

Entry

01
AUD $199

Cohort 1 price · Cohort 2 will be AUD $299

Learn the method. Build on your own time.

  • All 5 recorded sessions (30-day access)
  • Cohort Discord (30 days)
  • Founder Stack Vault
  • 15 Claude Routines and scheduled-task templates
  • 10 custom Claude Skills
  • AI Operating System setup guide
Recommended

Basic

02
AUD $299

Cohort 1 price · Cohort 2 will be AUD $449

Show up, build with the cohort, keep the network.

  • Everything in Entry
  • Live Zoom sessions with Arun
  • Live Q&A in every session
  • Live Day 5 automation build-along
  • Group routine / schedule setup
  • Shared troubleshooting board
  • Demo Day: product + automation walkthrough
  • All recordings + Vault updates
  • Private alumni Slack
  • 2× cohort office hours

Advanced

03
AUD $399

Cohort 1 price · Cohort 2 will be AUD $649

Live + a 1:1 with Arun the week you launch.

  • Everything in Basic
  • 30-min 1:1 launch review with Arun
  • Personal review of your product, offer, and Claude workflow
  • Custom Routine + Skill feedback
  • MCP safety review
  • Launch Day amplification to Arun's audience
  • Free retake in any future cohort
  • Curriculum updates as they ship

Cohort 2 will be AUD $299 / $449 / $649. This is the one time it’s cheaper.

§ 08 · The Vault

The Founder Stack Vault.

Everything we build together, pre-packaged for you to take and reuse in whatever you build next, and the one after.

Included in every tier

  • Prompt library50+
  • Claude Code Routine templates15
  • Cowork scheduled-task templates12
  • Custom Claude Skills10
  • Claude in Chrome workflow shortcuts5
  • CLAUDE.md starter templates
  • Next.js + Supabase + Clerk + Stripe starter repo
  • Supabase RLS review checklist
  • Stripe sandbox-to-live checklist
  • MCP safety checklist
  • Market brief / PRD / brand brief templates
  • Claude Design brief template
  • Launch swipe file
  • 30-day operating rhythm pack
  • Weekly competitor-monitoring workflow
  • Weekly SEO review workflow
  • Weekly product-health review workflow
  • Recorded bonus modules - API · Managed Agents · Agent SDK · Enterprise
§ 09 · The Instructor
Arun Prakash teaching a live workshop - a room of founders on laptops following along
Pl. 04 · Teaching, live

Taught by Arun Prakash.

Founder of orcalumi. Builds AI-first products, teaches live cohorts to operators, engineers, and founders.

Using Claude

Daily, since the API launched

Teaches

Live cohorts, in public + private

Background

Finance · Retail · Mining

I started orcalumi with a simple intent: to help companies with authentic visions go faster. The businesses I admire most are the ones solving real problems for real people. We’re not building a product empire. We’re here to help good companies accelerate and adapt in a world that’s changing fast.
— Arun Prakash · Founder, orcalumi

§ 10 · FAQ

Before you ask.

Anything else — write to hello@orcalumi.ai. Arun answers.

No. You need to be comfortable reading instructions, opening a terminal, and being patient with things that don't work the first time. Day 1 walks everyone through the Claude Code setup together - no assumptions. By Friday, non-technical founders in past private sessions have shipped working SaaS products on public URLs.

§ 11 · Not ready yet?

Join the list for the next cohort.

I’ll write when the next cohort opens, and occasionally between - a new Claude primitive, a case study, a rare discount. No drip sequences, no daily newsletter. Only when it’s worth reading.

Dates land here first.

§ 12 · The Last Page

Stop learning.
Start shipping.

Cohort 1 closes when 40 seats are full. Cohort 2 will be AUD $299 / $449 / $649 — and the stack will have moved again by then.

40 seats · from AUD $199

Takes ~3 minutes. Arun reads every one.