Field Notes: Real Patterns and Next Moves

This is Part 3 of a 3-part series. Part 1 covered the framework. Part 2 explored tactical approaches by role. This post shares what I’m actually seeing and doing.

---

I’m not an observer. I’m running these experiments in real time.

Here’s what’s actually happening.

The Patterns I’m Seeing

Pattern 1: The Full-Stack Intent Agency

Solo developers and micro-teams (2-3 people) delivering entire business verticals—code, strategy, GTM—in 4-6 weeks.

Charging: $300k-$1M per engagement

Replacing: $1.4M, 18-month traditional dev teams

Status: Early but real. I’m building this model right now.

Bubble vs. Build-Out: This model survives a correction. Why? Because it’s profitable from day one and delivers clear ROI.

Pattern 2: The Brownfield Wrapper

Enterprises using AI to map and wrap legacy systems instead of rewriting them.

Approach:

• AI maps the legacy codebase (90% automated)
• Humans validate business logic (10% manual)
• Modern API facade around old code
• New features in new stack, old features quarantined

Status: Happening at forward-thinking enterprises.

Bubble vs. Build-Out: This is pure build-out. It delivers cost savings and faster delivery. It survives—even thrives—in a correction.

Pattern 3: Architect-as-a-Service

Consultants shifting from selling process to selling judgment.

From: 200-person armies delivering PowerPoint

To: 1-3 person teams delivering production code

Status: The shift is happening. Consultants who can’t adapt are losing deals to smaller, faster teams.

Bubble vs. Build-Out: The old model depends on bubble money (inflated corporate budgets). The new model delivers ROI that survives a correction.

Pattern 4: The Utility Engineer

A new role: the person who manages the “last mile” of AI infrastructure.

Not building models. Not writing prompts. Building:

• Abstraction layers (multi-LLM provider routing)
• Monitoring systems (cost tracking, performance metrics)
• Fallback logic (when AI fails, what happens?)
• Security layers (validating AI outputs before production)

Status: Emerging. High-value role.

Bubble vs. Build-Out: Pure build-out. This is infrastructure work. It’s the “grid engineer” for the AI era. It survives market cycles.

Pattern 5: The Training Shift

Companies realizing they can’t “hire AI developers” fast enough. They need to retrain existing teams.

Why this works:

• Developers who know the business >> developers who know AI
• You can teach AI fluency in 8 weeks
• You can’t teach 10 years of domain knowledge

Status: Early but accelerating.

Bubble vs. Build-Out: This is hedging against both. If the bubble pops and hiring freezes, you’ve got trained internal talent. If the build-out accelerates, you’ve got capacity.

The Questions I’m Still Working Through

Let me be honest about what I don’t know:

Question 1: When Does the Correction Come?

Michael Burry is betting on a near-term correction. Maybe he’s right.

But timing market corrections is nearly impossible. The bubble could inflate for another 2 years. Or it could pop tomorrow.

I don’t know. Nobody does.

My Hedge: Build for profitability, not valuation. If the correction comes early, I survive. If it comes late, I capture more arbitrage.

Question 2: Does the Multiplier Hold Post-Correction?

If there’s a market correction and AI funding dries up:

• Do the tools get worse?
• Does compute get more expensive?
• Does the 50x multiplier collapse?

My bet: No. The infrastructure is already built. The models are already trained. The tools are already distributed. A stock market correction doesn’t make Claude Code stop working.

But I’m not certain.

Question 3: What Happens to Enterprise Adoption?

If CFOs start cutting budgets, do they:

• Cut “AI experiments” as speculative?
• OR double down on AI as cost reduction?

My bet: They do both. They cut speculative “AI strategy” consulting. They double down on proven productivity gains.

But again, I’m not certain.

Question 4: Is This Reproducible Across Skill Levels?

I got a 50x multiplier. But I’ve got 20+ years of experience.

Can a mid-level developer get the same results? Can a junior?

Is the multiplier about AI, or about the human directing it?

My bet: The multiplier is real but variance is high. Senior developers might get 50x. Mid-level might get 20x. Juniors might get 5-10x.

But I don’t have enough data yet.

Question 5: How Long Does the Arbitrage Last?

Right now, there’s a gap between what AI can do and what the market knows it can do.

That gap is the arbitrage.

Scenario A (Fast Correction): Stock market crashes in 12 months. AI hype dies. Arbitrage window extends because fewer people enter the market.

Scenario B (Slow Growth): Bubble inflates for 3 years. More people discover the multiplier. Arbitrage window closes faster.

I’m betting on 18-24 months of strong arbitrage regardless of scenario. But I could be wrong.

What I’m Actually Doing

Here’s my play. Not theory. Execution.

1. Building Cash Flow, Not Valuation

I’m not raising money. I’m not building for an exit. I’m delivering client work and stacking cash.

Why? Because if there’s a correction, cash is king. If there’s not, I didn’t need funding anyway.

2. Delivering Outcomes, Not Strategy

I’m not selling “AI consulting” or “AI strategy.”

I’m selling: “I’ll build your platform in 6 weeks for $500k. Production-ready. Deployed. Monitored. Guaranteed.”

That survives a correction. PowerPoint decks don’t.

3. Keeping the Team Tiny

I’m building a 2-3 person team. Not scaling to 50.

Why? Because:

• Lower overhead = higher margins = more resilient in a downturn
• Smaller team = less communication overhead = faster execution
• Quality over quantity = better outcomes = more referrals

4. Building Infrastructure, Not Hype

I’m building:

• Reusable code patterns (abstraction layers, monitoring systems)
• Documentation and knowledge transfer systems
• Training programs for clients
• Case studies with measurable ROI

This is infrastructure. It survives market cycles.

5. Documenting Everything

I’m writing about every experiment. Not to sell courses. To compress learning.

Because if the correction comes, the people who survive will be the ones who understood what they were building, not just riding the hype.

Where I’ve Landed—For Now

Here’s what I think today:

I believe:

• The 50x productivity multiplier is real and survives a market correction
• AI is infrastructure, not a speculative product
• A financial bubble and a technological build-out can coexist
• The old models (200-person consulting teams, 18-month dev cycles) are obsolete
• A new class of worker is emerging: Architects who deliver signal, not noise
• The arbitrage window is 18-24 months, but the transformation is 10+ years

I’m not betting on:

• Nvidia’s stock price staying at $3T
• VC funding staying easy
• Corporate budgets staying inflated
• Straight-line growth with no correction

My hedge:

Build for profitability, not valuation. Deliver outcomes, not process. Keep the team tiny. Stack cash. Build infrastructure that survives market cycles.

If the bubble pops, I survive because I’m profitable and I deliver real value.

If the bubble doesn’t pop, I capture the full arbitrage.

I’m betting with actions, not words.

Building the micro-team model. Taking the platform to market. Training others. Documenting everything.

If I’m right, I’ll have built something valuable. A new model. A new path.

If I’m wrong, I’ll have an expensive lesson in why you shouldn’t trust early numbers.

Either way, I’ll know. And I’ll write about it.

What This Means for You

I can’t tell you what to do. Your context is different. Your risk tolerance is different. Your opportunities are different.

But here’s what I’m trying—organized by role:

If you’re a solo dev or small team:

1. Pick a vertical where you can deliver measurable business value (not “AI consulting”)
2. Build the full stack (code + strategy + GTM) using AI-assisted development
3. Price on value delivered, not hours worked ($300k-$500k per engagement)
4. Build cash flow, not valuation (no VC, no hiring spree, profit from day one)
5. Move fast but build to survive (if the bubble pops in 18 months, you’re still profitable)

Timeline: 18-24 months of strong arbitrage. But build a business that survives 10 years.

If you’re an enterprise leader:

1. Map your brownfield with AI (weekend project, not 6-month discovery)
2. Wrap legacy systems, don’t rewrite them (modern API facade)
3. Retrain your team, don’t replace them (8-week program)
4. Measure value delivered, not hours worked (cycle time, cost reduction)
5. Build your correction story now (“while others speculated, we delivered ROI”)

Timeline: 12-24 months to establish AI-enabled delivery. This survives—and thrives—in a correction.

If you’re a consultant:

1. Shrink your team to 1-3 people (architects, not process-sellers)
2. Deliver working solutions in 6-12 weeks (not 3-year roadmaps)
3. Charge for outcomes, not hours ($300k-$1M fixed-scope engagements)
4. Build a portfolio of case studies with measurable ROI
5. If you can’t deliver in this model, you’re a parrot—adapt or die

Timeline: The correction kills the old model fast. Build the new model now.

If you’re a junior engineer:

1. Start as a junior architect, not a junior coder (systems thinking from day one)
2. Work with AI from day one (learn to direct it, validate it, review its outputs)
3. Pick a domain and go deep (become the person who knows both business and tech)
4. Build in public (document your learning, show your work)
5. Clear the bar or find a new path (this is harder than the old way, but higher leverage)

Timeline: 3-4 years to become a full architect. But you’re more valuable—and recession-proof—than traditional developers.

Where Do We Go From Here?

We stop asking “bubble or build-out?”

We recognize: Both. At the same time.

We stop waiting for clarity. Markets don’t provide clarity. They provide volatility.

We make intelligent bets:

• Bet on the build-out (real productivity, real ROI, real infrastructure)
• Hedge against the bubble (don’t depend on inflated valuations or easy money)
• Build to survive corrections (cash flow, not funding; outcomes, not hype)

We document. We experiment. We learn in public.

Because the people who capture value aren’t the ones who predict the future. They’re the ones who build robust strategies that work across multiple futures.

The infrastructure is here. The multipliers are real. The bubble might pop. The transformation continues anyway.

The only question left is: What are you building that survives both?

---

This post is part of the Signal Dispatch series on AI-enabled transformation. If you’re working through these questions—if you’re trying to figure out your own path through this—I’m happy to think through it with you. Reach out.

The experiments continue. I’ll report back.