What business structure should I choose for my startup?

Most venture-backed startups incorporate as a Delaware C-Corp for QSBS tax benefits and investor familiarity. Bootstrapped businesses often start as LLCs with optional S-Corp tax election.

What is the best bank for a startup?

Mercury is the default recommendation for most seed-stage startups — no fees, FDIC coverage up to $5M, and strong integrations. Brex is best when you need cards and banking combined. Bluevine works well for service businesses needing credit lines.

How can AI help startup founders?

AI serves as operational infrastructure for small teams: content pipeline automation, outreach personalization, meeting preparation, document synthesis, and financial tracking. It lets a two-person team compete with a twenty-person team.

What security should a new startup have?

Minimum security setup: password manager (1Password or Bitwarden), MFA on every service, YubiKey for email and banking, email aliases for service segmentation, and domain registrar lock.

Chapter 12

AI as Infrastructure

Most guides treat AI as a research assistant. That framing undervalues it by an order of magnitude. At a pre-revenue startup, AI is operational infrastructure. It is how a two-person team competes with a twenty-person team.

The reframe

The question is not "how can AI help me write things faster." The question is "which parts of my operation can be run by AI agents so that my human time is reserved for judgment, relationships, and decisions that actually require me."

The AI tool proliferation problem

The landscape is overwhelming. Cursor, Replit, Lovable, Claude Code, ChatGPT, Codex, Perplexity, Gemini — and that is just the shortlist. New tools launch weekly. Each one claims to be the one you need. The instinct is to try everything simultaneously. That instinct is wrong.

Here is the framework that actually works: pick a task, pick a tool, get good at it. Do not try to evaluate seven tools at once. Choose one problem — content drafting, code generation, research synthesis, financial modeling — and commit to one tool for that task for at least two weeks. Learn its strengths, its failure modes, and how to prompt it effectively.

Once you are proficient with one tool for one task, layer in the next. Over time you build a personal stack where each tool earns its place because you have tested it against real work — not because you read a comparison article.

Code & shipping

Tools like Lovable, Cursor, Replit, and Claude Code each excel at different scopes. Lovable is fastest for full-stack web apps from scratch. Cursor is strong for editing existing codebases. Pick based on what you are building today.

Research & analysis

Perplexity for real-time research with citations. ChatGPT and Claude for deep analysis and long-form synthesis. Gemini for multimodal tasks. Each has a sweet spot — none is best at everything.

Writing & content

Claude excels at nuanced, long-form writing. ChatGPT is faster for shorter drafts and brainstorming. Use one as your primary drafting tool and the other for second opinions on important pieces.

No long-term contracts

Most AI tools are monthly subscriptions. You do not have to commit to everything at once. Start with one or two paid tools, master them, and add more as your needs evolve. Cancel what you stop using.

AI prompts

Daily intelligence brief

You are my daily intelligence briefing system. My company is focused on [domain and thesis]. My target audience for content is [audience]. Here are today's news items from my source feeds: [paste items]. Produce a structured daily brief with: (1) the two or three items most relevant to my content pipeline, (2) any items that reinforce or challenge my investor thesis, and (3) one item I should send to a specific contact with a suggested one-line message.

Outreach personalization at scale

I need to send personalized first-touch messages to [number] prospects. For each contact I have: name, company name, company size, location, and one recent context item. Here is the list: [paste list]. Draft a personalized message for each. Tone: direct, no pitch, genuine. No more than 80 words per message.

AI tool stack audit

I am currently paying for these AI tools: [list tools and monthly costs]. My primary use cases are: [list tasks]. For each tool, tell me: am I using it for the right task, is there overlap I should eliminate, and what is the one tool I should master next based on my workflow gaps.

Agentic workflow design

I want to automate the following recurring task: [describe the task, inputs, desired output, and current manual steps]. I have access to: [list tools]. Design a workflow that automates this task end to end. Specify: the trigger, each step, which tool handles each step, and where human review is required.