This Is What Real Augmentation Looks Like

Joshua Wöhle had an idea for a new feature at a conference last month.

By the time he got back to his laptop, it was done.

Not "drafted." Not "half-finished." Done — because the moment the thought sparked, he blasted the instruction into his phone, a handful of agents went to work, and when he opened his machine later, the output was waiting for him. He didn't have to explain what he was working on, who the audience was, or what he'd already written on the topic. The system already knew.

That moment changed how I think about what AI augmentation actually is. (I teach this stuff. I thought I had a pretty sophisticated picture. I was undersold.)

The "100x Human" Problem

I've been semi-jokingly calling Josh the first 100x human I know. Semi-jokingly, because the number isn't the point. The stack is.

Here’s the setup. I’ve watched Josh and the Mindstone team up close for the past couple of years, and I’d had the same thought more than once: these people are operating in a way most of us still don’t have language for. So I asked Josh for receipts.

Josh and his team built a system called Rebel. He spends about four hours a day inside it. By his own rough estimate, he's about 100x more productive than he was three years ago.

When you actually try to decompose the 100x, though, it doesn't cleanly multiply. Something like 20x on engineering output, plus non-engineering work running in parallel, plus quality improvements, plus compounding memory. That's stacking, not multiplying — which Josh was the first to point out to me:

"A few multipliers stack: compression, parallelism, quality, and continuity. That gets much closer to the lived reality than a single before/after arithmetic claim."

That's the mental model most people are missing. And until I sat with Josh's note, I realized my own mental model was thinner than I thought.

Four Multipliers, Stacking

1. Compression: Same work, less time

Josh's first book chapter draft took about half a day. Pre-AI, roughly a week. His meeting prep workflow saves about 33 minutes per run. Rebel's own tracking showed 300 hours saved in the first two weeks.

Classic value capture. Same work, less time. This is the level most people stop at — and the level where most AI productivity writing stops too.

But compression alone is the on-ramp, not the destination. (See also: From Efficiency to Impossibility.)

2. Parallelism: Twenty-seven agents

This is where it gets interesting.

On one recent piece of work, Josh ran 27 agents in parallel on the same problem.

The bigger parallelism story is organizational. The main Rebel repository sits at over 7,000 commits, with more than 1,100 since March 1 alone. Humans, agents, and automation are all shipping against the same codebase.

And the point that matters most for the augmentation thesis: Josh doesn't have to choose between CEO work and product work. Product progress keeps happening while he does CEO work. Commercial work keeps happening too.

Josh pointed me to his CTO's public accounting: roughly 18x developer productivity — 400,000+ lines of code in about three months, versus a legacy app of similar size that had taken roughly five years. Josh's own caution is right: lines of code shouldn't be the centrepiece. But directionally, the point is hard to miss.

3. Quality: The trade-off that stopped existing

Here's what usually gets left out of AI productivity stories.

Pre-AI, Josh still personally handled about 85% of Mindstone's commercial output. (A thing growing-company CEOs don't love admitting out loud, but it's common.) There was always a trade-off — speed versus quality. Send more proposals, lower the bar. Keep the bar high, send fewer.

Now he produces fully personalized five-to-six page proposals in a few minutes, at a standard he says he previously couldn't reach consistently. Not "same quality, faster." Higher quality and faster. The old trade-off stopped existing.

The receipt: after our Florida AI Weekend, Josh sent twelve-plus personalized follow-ups in two days. Those produced two decision-maker leads and one investor lead. (The unassisted version of Josh sends three or four in that window, and they're thinner.)

4. Continuity: The compounding multiplier

This is the one almost nobody talks about, and it's the one I now think matters most.

Compression, parallelism, and quality each deliver a lift — but after that lift, you're running at a new steady state. Continuity is different. Continuity is the multiplier that keeps growing.

Here's what it means for Josh. His system preserves working context — decisions, relationships, open threads, what he's been writing on. Every conversation deposits something. Every future conversation draws from it. So when he blasts an instruction from his phone, the agents don't start from zero. They inherit context.

Josh, on how that feels in practice:

"You can have a thought that gets sparked in the middle of nowhere, just get the agent going, and when you're back at your laptop, that result is there for you to look at. … When you can just blast out orders wherever you are, and whenever you get back to your laptop they're all just done — this is a genuine next level of unlock, because otherwise you would have had to manage all of those things in parallel."

The usual price of parallelism is more management overhead — 27 agents should require 27 context loads. But with continuity, the context is already there. Parallelism gets dramatically cheaper to manage.

That's why continuity is the multiplier the other three compound under.

Models are getting cheaper and more interchangeable. The thing that actually compounds is context.

The Connecting Thread

Four multipliers. Each interesting on its own. But here's the actual unlock: Josh stays directly inside the product loop.

Commercial work keeps moving. Engineering keeps shipping. Follow-up happens at near-zero overhead. Memory compounds across all of it. And Josh is still making the judgment calls — what to build, who to talk to, which threads to pull on next.

In his own words:

"The real unlock is not 'I can do one thing faster.' It is 'I can now do things I previously would not have done at all — or not to that standard — while everything else is still moving.'"

Once you see it, you can't unsee the gap between most people's current AI use and this.

The Real Limit

One more line from Josh I keep coming back to:

"The limit is increasingly not model capability. It is my ability to imagine worthwhile questions."

That echoes something I wrote recently — the answer is yes, so what's your question? Josh is living the next-order version. When the stack works, your bottleneck isn't the tool. It's you — specifically, your imagination. And imagination is constrained by experience.

Which is why most people reading this won't suddenly become Josh. They don't yet have a visceral model for what this feels like, so they can't picture the question they'd even ask.

What to Actually Do

1. Stop asking "how do I save 20 minutes?" Start asking "what could keep moving while I'm elsewhere?"
That question points at parallelism — the multiplier almost everyone skips. Saving time is 101. Having things happen while you're in another meeting is 201.

2. Name your augmentation backlog.
Open a document. List three things that went undone last week because you were the bottleneck — proposals you didn't send, follow-ups you didn't write, ideas you didn't explore. Don't solve them yet. Just name them. That list is the map.

3. Start a context document for one project.
Pick one thing you work on repeatedly — a client, a product, a book. Keep a running document with decisions, constraints, voice, history. Feed your system from it. Six months of this, and you'll have the beginning of what Josh has: a system that already knows.

The Opportunity Cost

Josh has one more line I think about constantly:

"The opportunity cost of being an unassisted human is now much higher than ever."

He means it operationally. I mean it existentially. The gap between an unassisted operator and a Josh-style operator isn't a rounding error anymore. It's a chasm — and it widens every week, because every week the augmented operator's memory gets bigger.

Serious question: if you could do in four hours what most people do in eight — while work kept moving in the background — what would you build with the other four?

That's the real question. And it's not about Josh.

Josh had a product idea at a conference. By the time he got back to his laptop, it was done. What thought did you have this week that's still waiting on an unaugmented version of you?

Related: From Efficiency to Impossibility: How the Trail Blazers Prove AI Creates Value
Related: The Answer Is Yes. What's Your Question?
Related: It's a Skill, Not a Pill

Join over 34,147 creators & leaders who read Methods of the Masters each week.