What I’m thinking about: What it actually looks like when AI stops being a novelty and starts replacing full work days…in a single week.
Returning readers already know how bullish I am on AI (tons of past examples).
But something shifted over the ~4-6 weeks. The capabilities of LLMs, and specifically Claude Cowork, are just leaps and bounds beyond anything we were experiencing even three or four months ago, let alone a year.
I’m not exaggerating when I say that tactics we were using two weeks ago in our business are already outdated…and the ground keeps shifting.
As an analogy, over the past few decades, it generally took between 3-4 years for new video game consoles to be released (e.g. N64 to GameCube, or PS2 to PS3). And fairly linear jumps in regard to graphics or processing power.
With AI, as good as the models were getting less than a year ago, there were still recurring hallucinations, limited context windows, clunky long-form outputs, requirements to ‘prompt engineer’, and unreliable connections to other software.
Compared to last summer, today’s AI models are like we skipped from the original NES…straight to the PS5. The generation gaps are collapsing on top of each other (and getting faster by the week).
Let me lay out what we recently accomplished…in less than a week. Specific examples. Real time savings. Hard ROI.
=====
Quick Context on Claude Cowork
If you haven’t used it yet, Cowork essentially merges the conversational simplicity of a standard LLM with the more powerful coding and file-processing capabilities of Claude Code or ChatGPT Codex…without needing any development experience.
Think of it like an assistant with near unlimited potential to deliver WHATEVER you ask for (if this sounds too good to be true, keep reading.)
It lives locally on your desktop (Mac or Windows). We upgraded to the $100/month Max plan because you’ll burn through the $20/month tier pretty fast (but you can get started at $20/month to be clear).
My team and I, and the best AI practitioners I know (e.g. Callan and her UnCommon Business team), LIVE inside Cowork now.
Onto the examples:
A 40-Page Crypto Audit Report…in Under 30 Minutes
Some of you read about my recent Helium crypto theft. (Entire ~$50K+ position stolen by a sophisticated thief, right under my nose…a year before I noticed. Sub-5% chance of ever recovering anything.)
Given the amount of capital, I wanted to explore every possible avenue.
The crypto exchange where the thief offloaded the tokens was at least responsive when I reached out…however, they required law enforcement to submit an official request. So I needed a detailed report to file with police.
So, I took ~15min to screenshot the entire flow of funds across the blockchain (one of the benefits of crypto…you can follow where tokens go), dumped all the screenshots into a folder without even labeling them, and gave Cowork access to that folder.
Summary of my prompt to Cowork: “You’re a top crypto investigator, here are the screenshots, I need a comprehensive audit report suitable for law enforcement to pursue recovery through this exchange.”
In less than 30min, it produced a nearly 40-page report. Color coded. Exec summary. Precise timestamps of the theft. Key wallet addresses. All supporting screenshots and evidence mapped out in a logical chain of custody.
95% accurate on the first pass.
A couple formatting issues and one error on the total amount sent to the exchange (always double check AI’s work…ALWAYS), but overall…stunning.
That report would have been 8-10+ hours of my time to produce manually. Certainly wouldn’t have looked as clean either. And I got to stay focused on SLC instead of getting buried in crypto forensics.
(Do the math on likely return of any time/money investment. Was a sub-5% chance of recovery worth a full day’s work? Questionable. But for less than an hour total, and most of that time without my direct execution? Certainly.)
I sent the report to the Austin Police Department (APD) and the FBI. A detective got back to me quickly and complimented how organized it was. His feedback was that most people who submit these are way less organized…and he said he’s never seen a better quality audit report (+1 to Cowork).
To tie a bow on this story: APD did a ton of validation on the theft, and even sent in a subpoena to freeze an entire other exchange with millions of dollars of daily flows, that was part of the thief’s transaction history.
Ultimately, as expected, it was determined my tokens were cashed out immediately and lost within a convoluted history of future trades. Law enforcement’s hands are tied, it’s a dead end. At least the matter is resolved, letting me move on mentally, and giving me an excellent practice rep with Cowork.
(As a reminder, think about your counterparty on the other end. If you give people a clear roadmap on what to follow, they’re way more likely to help you out. That applies to law enforcement, investors, partners…everyone.)
=====
Newsletter Prompt Optimization
We’ve published almost a year’s worth of Serious News newsletters at this point.
I wanted to analyze which subject lines and preview text copy actually drove the best open rates.
So I just asked Cowork. Positioned the problem, let it determine the right metrics to pull, fed it the data, and went back and forth a bit on the findings.
It identified the top performing patterns (mindful that we don’t have thousands of data points yet, but enough to start building a trend) and updated the underlying newsletter prompt to account for those findings.
As a bonus…it also flagged repetitive and superfluous language within the prompt itself. Cleaned it up to reduce token waste when the AI processes it. (Cowork recognized that this would be an optimal step within the process and suggested it to me without my direct request.)
And Opus 4.6 (Claude’s most powerful model) specifically has gotten noticeably better at maintaining my voice when I dictate raw thoughts and let the prompt handle formatting.
The annoying part of writing has always been the formatting…paragraph breaks, bold/italic placement, getting it into the delivery tool. That’s the “Ugh, I don’t wanna do this,” inertia that leads to procrastination.
But when I can just feed my thoughts and ideas in and not worry about any of that? It makes me want to write even more.
=====
150+ Pages of Legal Documentation…in One Shot
This one almost broke Cowork!
We’re finalizing the handoff of Land Pricer to a new operating team (new CEO, our former CTO, targeting enterprise sales to appraiser firms).
The legal complexity here was significant:
An existing LLC assigning IP to a new entity. Different investing rights. My existing equity transferring into the new company. New incentive structures for the incoming CEO and CTO. Anticipated design partners who’d be both customers AND potential equity holders. Plus terminating the old LLC language that rewarded the CTO in a different format.
All of it had to be addressed properly, and a full attorney office prep of these docs would be at least $10K.
Previously, I would have gone back to the ChatGPT legal project I used for a recent entitlement deal LLC (which took almost two full days of my focus to produce). But instead, I fed Cowork every relevant legal document I had.
The existing Land Pricer LLC. Our redlined rubric for the new agreement. Several other attorney-drafted legal docs that could serve as boilerplate language. DEEP context in my request.
(This is critical…over-index on giving AI proper context. Pro tip: Use audio transcription to speak to the AI models instead of typing. The output from LLM’s will always be substantially better with more context, regardless of how good the underlying models become.)
Cowork produced 10 separate legal documents…including ones I didn’t even ask for.
One was a summary document, as if I were receiving these from a law firm. Highlighting the most critical points, what’s most time sensitive, what’s easier to negotiate versus what to hold firm on, and all the rationale.
Over 150 pages across all 10 docs. Some 25-30+ pages individually.
It stalled out a handful of times during the process. Had to restart it a couple times as it bugged out on me (Cowork told me this task was a major ask). A couple docs had formatting errors when opening in Word, which were quickly fixed.
A couple hours with Cowork (again, most of it working in the background without my input), and we had legal-ready company docs for final team review and signatures.
=====
SEO/GEO Output: Less Time, 5X the Articles
Just over a month ago, we did a time audit on every repeatable task in the business. Used Loom recordings plus an AI analysis prompt to determine where the biggest time savings lived, and the lowest-hanging fruit for automation targets.
One of our biggest bottlenecks was SEO and GEO (optimizing to show up within AI search results like ChatGPT, Perplexity, Gemini) article production for SLC…and SEO taking almost 58 minutes per article specifically.
(FYI these articles drive our #1 marketing mechanism in terms of ROI.)
Without us even implementing any of the suggested workflow automations (e.g. Zapier), Claude’s base model improvements a couple of weeks after our time audit dropped SEO article production to 17 minutes.
And instead of being satisfied, we applied the Alex Hormozi principle of asking, “Can we do MORE?” Take the time savings and multiply output.
So I asked my team to 5X our previous output…to 5 SEO and 10 GEO articles per week.
Within 3 business days (the goal was within 5 days), our team accomplished this through Cowork, plus additional refinements cut the time per article from 17 minutes to 5 minutes…for the ENTIRE batch.
This represents a 98%+ increase in speed, alongside 5X the output, compared to where we were at just over a month ago.
(As an extra bonus, we can schedule tasks within Cowork, so these articles are automatically prepared and ready for review/publishing when we log in on Monday.)
=====
$1.1M Deal, Investor Deck Ready in Hours
We recently reviewed a large deal that would require external capital to commit.
Previously, building a sophisticated investor deck summary for a raise took ~3-4 days of iteration back and forth internally. Now with Cowork, feeding it examples from previous raises plus the new deal info…we had an investor-ready deck within a couple hours.
This is only going to become more important as we pursue upmarket deals with more capital partner involvement, and allows us to move even faster on hot, competitive opportunities.
=====
Even Outside of Land
My wife, a clinical therapist, had been using a Claude project to help format patient notes…but one patient at a time. If she was seeing multiple patients per day (like every healthcare provider out there), the backlog would build and the stress would mount.
With Cowork, she can now parallel process ALL of her notes simultaneously and just do a final review, erasing the recurring backlog build-up issue overnight.
=====
As a reminder, all of the above happened within a SINGLE week, and we’re just getting started with realizing the full potential of Cowork.
This is the most excited I’ve ever been as a professional. Regardless of (terrible) market conditions. Regardless of turmoil. It honestly feels like having a magic wand in hand.
If you haven’t tested Cowork yet, or you’ve been sitting on the sidelines with AI in general…NOW is the time to get moving. We’ve passed the Rubicon and business (and our lives) will NEVER be the same.
=====
From the above examples, the time savings from AI aren’t theoretical anymore. They’re real, and they’re compounding…daily. If you’re an experienced operator with routine deal flow and you want a capital partner who’s leveraging every possible edge to protect the downside (and maximize upside) in the nastiest market in decades, send us your best deals.
We write checks from $50K to $1M+. We close 100% of deals we commit to. And we bring national underwriting experience to every transaction.
