By the time a typical "AI for creators" cohort ships, 30-60% of its tool-specific training is obsolete: Beehiiv pushed the MCP server in March 2026 and the Beehiiv lesson needs a rewrite; Lovable added $100M ARR in a single month and the founder-sell pricing shifts; Notion AI Agent 3.0 ships and the Notion lesson is half-obsolete. Tools have a six-month half-life. Categories have a ten-year one. The L1 operator who organizes their stack around seven enduring jobs-to-be-done - research, drafting, audio/video, distribution, monetization, community, build - has an operational anchor that survives any one tool getting acquired, repriced, or shipping a breaking change. This lesson is that map, with the May 2026 canonical tools per category named for orientation, not worship - because the canonicals in 2027 will be different and your category framework will not need updating.

Why Categories Before Tools (The Anti-Vendor-Lock Insurance)

If you go to a typical "AI for creators" cohort in 2026, the curriculum is structured around tools: a lesson on Beehiiv, a lesson on Castmagic, a lesson on Lovable, a lesson on Notion AI. The implicit promise is that the cohort is a tour of the current best-in-class. The problem with that structure is its half-life. Notion AI Agent 3.0 ships and the Notion lesson is half-obsolete. Beehiiv pushes the MCP server in March 2026 and the Beehiiv lesson needs a rewrite. Lovable adds $100M in ARR in a single month and the founder-sell pricing shifts. By the time you finish a tool-first cohort, you have somewhere between 30% and 60% obsolete training.

The fix is category-first. Tools change; the jobs-to-be-done change much more slowly. Research is research; drafting is drafting; community is community. The 2026 winners in each category will shift by 2027, but the existence of "you need a research tool" doesn't. So we teach the categories, name-check the current tools as canonical examples, and update annually. The operator who maps their workflow against the seven categories has a stack that's debuggable: when a category is empty (no research tool, no brand-memory store), the gap is visible; when a category has redundancy (three brand-memory tools), the cancellation is obvious.

Category 1: Research and Brand Memory

The category that feeds every other category. Two sub-jobs combine here: research (find facts, citations, source-grounded answers) and brand memory (store your voice corpus, past work, audience-signal inputs in one substrate).

Research (2026 canonical tools)

• Perplexity - Search-first AI with named citations. The default for "find me a source for this claim."
• NotebookLM (Google) - Up to 50 sources, 500K words each; retrieval-based; the back-catalog research tool.
• ChatGPT Search / Claude Web Search - Grounded search inside your drafting tool. Convenience matters; switching tools costs time.
• Reka, Exa - Newer entrants for specialized research.

Brand Memory (2026 canonical tools)

• Notion AI Agent 3.0+ - The safe default. Personal knowledge bases + workflows + AI in one substrate.
• Mem - The AI-native option; strongest auto-organization.
• Reflect - The calmer-UI option; daily-notes-first.
• Granola - If your input is mostly meeting transcripts (calls, interviews); auto-summary at the surface.

Persistent Context (Claude Projects, Custom GPTs, Gemini Gems)

The other half of brand memory: where your voice corpus, system prompts, and per-task rules live. Most operators pick one (Claude Project is the L1 default for voice work; Custom GPT for orchestration-heavy work).

L1 rule for this category: one research tool + one brand-memory substrate + one persistent-context home. Three tools, total spend ~$40-$60/mo. Most operators in 2026 have 3-4 redundant brand-memory tools running simultaneously, which is the most common L1 stack-audit finding.

Category 2: Drafting (Text)

The workhorse category. This is where Tuesday issues, video scripts, LinkedIn posts, course modules, and support replies actually get written. Picked carefully (lesson 1.3), one tool here is enough at L1.

Drafting / Text (2026 canonical tools)

• Claude (Opus 4.5) - Best voice consistency across long outputs. L1 default for the voice-conscious creator.
• ChatGPT (GPT-5.2) - Broadest plug-in / connector / MCP ecosystem. L1 default for the orchestration-heavy creator.
• Gemini (3) - Longest effective context (1M effective vs. 2M advertised). L1 default for back-catalog query.

In-Tool Newsletter AI

• Beehiiv AI suite - AI Writer (drafting), AI Image (inline visuals), AI Analytics (post-send retro), plus the Beehiiv MCP server (March 2026 release; V1 read-only, V2 write capabilities announced) - lets you operate Beehiiv from inside Claude, ChatGPT, Gemini, or Perplexity.
• Kit (formerly ConvertKit) AI - Email drafting, subject-line variants, commerce.
• Substack AI - Limited drafting features; Substack is more platform than AI vendor.

Social and Thread Drafting

• Typefully, Hypefury, Tweet Hunter - X/Twitter scheduling + AI variants.
• BoringEmail, Shortform - Short-form newsletter / digest drafting.
• Gamma, Tome - Slide / deck drafting from prompts.

L1 rule for this category: one primary text drafter (Claude / ChatGPT / Gemini). Add an in-tool layer only if your newsletter platform's AI is materially useful (Beehiiv's MCP integration makes it so for Beehiiv operators specifically).

Category 3: Audio and Video

The category where 2025-2026 quality jumps were largest. The full stack for a working creator:

Recording and Editing

• Descript - Text-based video and audio editing. The highest-leverage solo-creator tool in this category.
• Riverside - Browser-based multi-track recording. Up to 4K video and lossless audio for guest interviews.
• Buzzsprout AI - Hosting + AI features (transcription, summaries, social clips).

Repurposing (Podcast and Video → Short-Form)

• Castmagic - One episode → newsletter section + clips + threads + LinkedIn + show notes + quote graphics + Pinterest pins. ~$120K MRR as of 2026; canonical example of the repurposing category.
• Wisecut, Pictory - Alternative repurposing engines.
• Opus Clip - Long video → short-form clip auto-selection with hook detection.
• Submagic, Captions, Veed, InVideo - Caption and polish for short-form.

Voice and Music

• ElevenLabs - Voice cloning, Voice Isolator, ElevenMusic (April 2026 launch), ConvoAI (agent voices).
• OpenAI Voice - Bundled with ChatGPT.
• Suno (v4.5), Udio - Music generation (note unsettled training-data lawsuits).

Synthetic Talking-Head and Video

• HeyGen, Hedra, Synthesia - Avatar / talking-head synthesis. Hedra (2026) crossed quality threshold for short Reels and ads with disclosure.
• Tella, Loom AI - Async screen recording + AI cleanup.
• Veo, Sora - Generative b-roll and short video.

L1 rule for this category: Descript first (transcript-based editing is the highest-leverage feature). Add Castmagic if you publish long-form regularly. Add ElevenLabs Voice Isolator if audio quality is a bottleneck. Defer HeyGen / Synthesia until L4-or-later use cases.

Category 4: Distribution

The category most creators under-invest in. You shipped the issue; now you have to push it across surfaces.

Owned Distribution Surfaces

• Substack Notes - Highest organic reach for written content per same-content A/B in 2026.
• Threads (~320M MAU, fastest growing of the major text-social).
• X / Twitter, LinkedIn, Bluesky (~35M MAU).
• YouTube, TikTok, Instagram - for video repurposing.

Scheduling and Queue Management

• Typefully, Hypefury - X/Twitter scheduling + thread management.
• Buffer, ShortStack - Multi-platform scheduling.

Newsletter Cross-Promotion

• Beehiiv Boosts, Kit Sparkloop, Refind, SparkLoop Upscribe - Pay-per-subscriber acquisition via cross-promotion.

L1 rule for this category: pick one scheduling tool. Pick one cross-promotion network if your newsletter is paid-ready. The L2 Ch5 lesson formalizes the cross-platform repurposing matrix.

Category 5: Monetization

The category that funds the rest. Tool selection here is partly platform decision (Beehiiv vs. Kit vs. Substack) and partly checkout/payment infrastructure.

Newsletter Monetization Platforms

• Beehiiv - Paid tiers, ads, digital products, $99/mo Max. Flat fee, no revenue share.
• Kit - Commerce features, $33/mo Creator plan, full commerce stack.
• Substack - 10% revenue cut, strong network for discovery, weaker commerce features.

Course and Cohort Platforms

• Maven - Live cohort courses; 96% Week 1 → Week 2 retention per a16z. 10% revenue cut.
• Teachable, Kajabi, Thinkific, Podia - Self-paced course hosting.

Info-Product and Checkout

• Gumroad, Lemon Squeezy, Stripe Checkout, Polar (Merchant of Record), Whop (2.7-3% fees).

Entity and Payment Infrastructure

• Stripe Atlas - LLC / S-Corp incorporation for solo creators.

L1 rule for this category: pick your newsletter platform deliberately (Beehiiv vs. Kit vs. Substack - covered in L4 Ch4.2). Pick one checkout. Defer the entity decision to L4 Ch7.

Category 6: Community

The recurring-revenue surface most creators delay too long.

Paid Community Platforms

• Circle - $89-$360/mo, premium branding, white-label feel.
• Skool - $9-$99/mo plus 2.9% transaction; gamified; 180K+ free members in the Skoolers community.
• Discord - Free-ish, with Whop Communities at 2.7-3% transaction for paid.

Community Automation

• Circle AI - In-platform AI for prompts, moderation, welcome.
• Custom Claude / GPT support agents - Inbox triage and welcome cadence drafting.

Solo Support Ops

• Intercom Fin, Help Scout - Customer support inbox tools with AI features.

L1 rule for this category: identify whether community is part of your offer ladder before tooling. The L4 Ch2.3 lesson covers the Skool vs. Circle vs. Discord-plus-Whop decision in depth.

Category 7: Build (Indie SaaS / Micro-Product)

The newest category for creators in 2026. Two years ago, "creator" and "indie SaaS founder" were distinct identities. By 2026, the line has dissolved - Lovable's $400-500M ARR by May 2026 (TechCrunch confirmed crossing $400M in February, adding $100M in a month with 146 employees) is largely creators shipping micro-products to their newsletter lists.

AI App Builders (2026)

• Lovable - Roughly $400-500M ARR by May 2026; non-developer-friendly.
• Bolt - $40M ARR in 4.5 months; rapid prototyping.
• Replit Agent - ~$250M ARR, $9B valuation; broader engineering platform with creator-friendly mode.
• v0, Base44, Mocha - Adjacent options.

Back-Ends and Deployment

• Supabase, Neon, PlanetScale - Databases.
• Vercel, Netlify, Railway - Deployment.

Payments and Checkout (Build-Specific)

• Stripe, Lemon Squeezy, Polar (Merchant of Record for global VAT), Paddle.

L1 rule for this category: the build category is optional at L1. It enters operationally at L5 Ch2 ("From Newsletter Reply to a $19/mo Tool in 14 Days"). At L1, just know the category exists and the canonical 2026 tools so you don't over-build a 2025-era stack.

Seven Categories: 2026 L1 Picks at a Glance

Decision rule: Use one tool per filled category; mark categories you have deliberately deferred. Total L1 spend should land $80-$150/mo. Any month you exceed $200, run the redundancy audit.

Composite Case A: Sara the Course Creator

Composite, drawn from L1 stack-audit findings across three operator coaching sessions in February 2026. Sara runs a $497 design-systems cohort course (4 cohorts/year, ~28 students per cohort) and a free 6,200-subscriber newsletter. Pre-audit stack: Notion AI ($10), Mem ($15), Reflect ($10), Granola ($14) - four brand-memory tools, none used as the primary; ChatGPT Plus ($20), Claude Pro ($20), Gemini Advanced ($20) - three drafters running in parallel with no clear primary; Descript ($24), Riverside ($24), Loom AI ($15) - three overlapping video tools; plus Perplexity ($20). Monthly spend: $192. Stack-map exercise revealed four empty boxes (no scheduler, no community tool, no monetization tool beyond Stripe, no build) and seven boxes with redundancy. Decisions: kept Notion AI ($10) as the sole brand-memory; kept Claude Pro ($20) as sole drafter; kept Descript ($24); added Castmagic ($39) for the repurposing gap. New total: $93/mo. Net savings: $99/mo = $1,188/year. Recovered tool-juggling time: roughly 4 hours/week (one full publish-day per month). Six weeks later, she had shipped two new module-launch sequences she previously had no time for; cohort 4 closed at 36 students instead of 28, an $3,976 revenue lift attributable to recovered capacity.

The Most Common Failure Mode

The most common stack-construction failure is filling the categories you find easy and ignoring the ones you find intimidating. Newsletter operators over-stack research and drafting (two of each) and leave monetization at "I'll get to Beehiiv eventually." YouTubers over-stack video tools and have no brand-memory substrate, which means their voice corpus lives in seven Google Docs. Indie SaaS founders pile up Lovable + Bolt + Replit but have no research tool, so they ship product hallucinations into customer-facing copy. The fix is mechanical: map all seven categories first, identify which are empty, then fill the empty ones before adding redundancy to filled ones. The most expensive line in any operator's monthly spend is usually the third tool in an already-filled category. The most valuable next purchase is usually the first tool in an empty one. Audit quarterly. Cancel the redundancies in the same hour you identify them - calendar-confirmed, not "I'll get to it."

Week 1, Week 4, Week 12: Stack Maturity

Week 1. You complete the seven-category mapping. You discover two redundancies and one missing category. You cancel one tool that hour and add one new tool by end of week. Net spend changes by less than $20/mo.

Week 4. You have run four publish cycles with the new stack. The added category (often repurposing or scheduling) is paying for itself in recovered hours. You cancel a second redundancy.

Week 12. Total stack is $80-$150/mo; every paid tool gets used at least weekly; all seven categories are either filled or marked as deliberately deferred. Quarterly audit becomes a 15-minute check rather than a stressful rebuild.

The Modality Overlay (How the Seven Map to L1 Ch1.3)

The seven categories overlap with the four modalities from L1 Ch1.3 (text, voice, video, image). Quick map:

• Research and Brand Memory - primarily text modality.
• Drafting - primarily text modality.
• Audio and Video - voice and video modalities; some text for transcripts and show notes.
• Distribution - text-heavy for written platforms, video-heavy for YouTube/TikTok.
• Monetization - modality-agnostic; tooling rather than content.
• Community - text-heavy at L1; voice and video at L4-L5 for office hours, group video calls.
• Build - modality-agnostic at the tool level; modality-specific at the product level (a Lovable-built voice transcription tool is voice).

The two frames (categories and modalities) together let you map your stack on two axes: what job is this tool doing (category) and what kind of content is it producing (modality). Most stack-audit findings sit at the intersection of "redundant category" (three brand-memory tools) and "underused modality" (paying for image gen but never publishing visuals).

The L1 Deliverable: The Seven-Category Stack Mapping

Open your one-page L1 doc. Add a new section called Stack Map. List your current AI tools and tag each with its category. The seven categories should each have one tool or one explicit gap. The two most common patterns from L1 audits:

1. Redundant brand-memory. Notion AI + Mem + Reflect + Granola all running. Pick one; recover $40-$80/mo.
2. Missing repurposing layer. Long-form creator with no Castmagic / Opus Clip / Submagic; doing repurposing by hand. Add one tool; recover 5-10 hours/week.

These two findings, alone, are typically worth $1,500-$3,000 in annual cost recovery and time-back at common operator scales. The L1 capstone "AI Stack Audit + One Shipped Fix" is built around this mapping.

Why This Matters for L4 Strategic Decisions

The seven-category mapping is not L1-only. It becomes the analytical backbone for the L4 Ch4 stack audit (12 tools to 4), L4 Ch4.3 ghost-team operating system (each category gets a named "agent"), and L5 Ch1.1 ghost-team org chart. Get the categories right at L1 and the L4 work plugs in cleanly. Skip them at L1 and the L4 stack audit feels like starting over.

"Categories endure; tools change. Mapping by category instead of by brand is what lets you swap Notion AI for whatever ships in 18 months without rewriting your operating system."

Key Takeaways

• The 2026 creator-AI ecosystem organizes into seven categories: research and brand memory, drafting, audio/video, distribution, monetization, community, and build.
• Categories endure; tools change. Category-first thinking is anti-vendor-lock insurance.
• Each category has canonical 2026 tools - Perplexity / NotebookLM (research), Claude / ChatGPT / Gemini (drafting), Descript / Castmagic / ElevenLabs (audio/video), Typefully / Hypefury (distribution), Beehiiv / Kit / Substack (monetization), Circle / Skool (community), Lovable / Bolt / Replit (build).
• L1 rule: one tool per category as the default; defer multi-tool until L3-L4 when scale justifies.
• Most common L1 audit findings: redundant brand-memory (3-4 tools where 1 fits) and missing repurposing layer.
• The seven categories overlap with the four modalities (text/voice/video/image) - the two frames together let you map a stack on two axes.
• L1 deliverable: a Stack Map section in the L1 doc tagging each current tool with its category; identifies redundancies and gaps.
• The seven-category framing is the analytical backbone for L4 Ch4 (stack audit) and L4 Ch4.3 (ghost-team OS) - getting categories right at L1 makes L4 work click.