The end of “make-work” knowledge work
David Peterson

There are some weeks where the universe seems to be humming a single note, and if you’re quiet enough, you can hear it. I had one of those weeks recently, capped by two walking meetings on a Friday that never should have rhymed, yet sang in perfect, eerie harmony.

The first meeting was with the head of research at a hedge fund, who is trying to unearth non-obvious insights from non-obvious datasets. The second meeting was with one of my oldest friends who is now the sole software engineer at a next-gen battery manufacturing startup, tasked with building the entire software layer for their first factory.

Strangely enough, our conversations ended up being about the same topic.

"There used to be a lot of hard stuff," the head of research told me, "a lot of effort just structuring unstructured data." The factory engineer echoed her: "You used to have to put so much work into just getting data off the machinery and making it understandable."

But most of that just isn’t that hard anymore. Advances in IoT have made machinery data easily accessible. LLMs have made extracting insight from complex documents child’s play.

"Now the bottleneck," they both said in their own way, "is asking the right questions."

I didn’t really believe them at first, I’ll be honest. Surely there is loads of software that needs to be built. Surely there are still many problems to be solved?

And yet, when pushed, not really. Indeed, here’s an exact quote from my factory engineer friend: “I'm trying to get rid of the traditional software - MES, WMS, whatever else. Just go pure data lake + predictable API on top + LLM interface. And it’s kind of working??”

These aren’t AI bulls, mind you. She doesn’t run research at an AI-first hedge fund. He doesn’t run engineering at a first-of-an-AI-kind factory. They’re both just professionals. Doing their jobs. And realizing how much the game has changed.

And that’s what I want to talk about today. I think we’re witnessing the end of “make-work” knowledge work and the cold realization that we aren’t so good at the work that really matters.

For the last two decades, a huge portion of what we called "knowledge work" was really just digital plumbing. It was the mundane, though often well-paid, but still mind-numbing, labor of sweating the middle layers of the software stack. It was writing PDF scrapers or building custom OCR models to extract data and cram it into rows and columns. It was massaging CSVs in Airflow or dbt. It was wrangling massive SQL queries to generate custom charts and dashboards.

That era is over. The sort of work most of us have spent our entire careers doing is now automated, cheap, or both. Which leaves us all to focus on the one thing that’s left (and, in reality, the only thing that ever mattered): asking good questions.

And indeed, this is what that hedge fund researcher was actually most concerned about. 20 years ago when she got started in the business, it took so long to build any sort of novel data set, that you had to be sure it was a worthwhile endeavor. You’d never wonder whether it was a good question to ask because that was a prerequisite.

Now she’s wondering if her analysts even know how to ask interesting, novel questions anymore. Not because they aren’t smart. But because when you can ask anything, it’s hard to know where to start.

There’s a cruel irony here. Just as the ability to think deeply and ask good questions becomes the only durable human advantage left, generative AI is systematically draining our capacity to do it.

The cell phone erased our memory for phone numbers. GPS wiped out our mental maps. Infinite scroll nuked our attention spans.

Now, generative AI offers to synthesize every thought before we can fully form it. Why wrestle with a blank page when a model can suggest headlines, outlines, and counter-arguments in milliseconds? We are eagerly outsourcing the very faculty (slow, effortful, disciplined cognition!) that the new economy suddenly prizes above all else.

I’m not sure if I fully believe that long-standing systems of record can be so readily replaced by pipes, storage, LLMs and a conversational interface (though I will report back on my friend’s journey!), but I am fully convinced that knowledge work is changing. Soon enough, the only job left will be thinking, but what if we forget how?

No More Painting by Numbers
It’s the end of the “SaaS playbook.

The Age of Artisanal Software May Finally be Over
Every wave of technological innovation has been catalyzed by the cost of something expensive trending to zero. Now that’s happening to software.

Founders as Experiment Designers
David on why founders should run everything as an experiment.

When Growth Stalls
Or why to kickstart growth you should narrow your ICP.

ENTERPRISE/TECH NEWS

Browser Wars - OpenAI are reportedly developing their own browser built on Chromium and rumored to include a native ChatGPT interface plus the "Operator" agent that can perform tasks like filling forms and booking reservations. This move is seen as a direct challenge to Google's Chrome, which currently dominates with over two-thirds of the global browser market. TBPN, on their daily show, discussed that other AI-native browsers like Perplexity’s Comet, Brave, Edge with Copilot, and The Browser Company’s Dia are also part of what's being dubbed "Browser 2.0," marking a major shift toward conversational, agent-driven web experiences. “This is the oldest plan in tech: find product‑market fit with a single killer use case, then vertically and horizontally expand until you control the interface layer itself… Welcome to the next generation of browser wars.”

HOW TO STARTUP

Blitzhire - Villi Iltchev’s particularly topical and useful breakdown of the new acquisition trend which he calls “Blitzhire” is worth reading this week. “Windsurf, Character.AI, Inflection, ScaleAI, and Adept are all examples of very exciting AI startups that have forgone the traditional M&A process and adopted a new structure that basically transfers the employees and perhaps the IP to the buyer.” These aren’t typical acqui-hires or outright buyouts; the structure transfers key talent and tech instantly, while leaving the original startup alive and well. It’s messy, tax-inefficient, legally complex, and while it’s unlikely to avoid regulatory scrutiny for long, in the race for AI dominance, for now speed trumps neatness. 

HOW TO VENTURE

No Comment - This week Sequoia Capital took the view that sometimes silence is golden, when their Partner Shaun Maguire courted recent controversy. Techcrunch wrote about the aftermath of Maguire’s tweets, where although he addressed the controversy himself via a 30 minute video, Sequoia themselves made no comment. The article cites some opinions that the controversy may even drive deal flow for the fund, but also quotes an unnamed crisis communications professional who ‘has some advice for Maguire and, by extension, Sequoia.’ “Regarding the video Maguire published in the aftermath of his initial comments, the expert said, “I did think that apology addressed the ambiguities in [Maguire’s] post. But it’s a 30-minute video — you have to be really interested to watch this.”

What the Dickens - In Upstarts Alex Konrad writes about the ‘tale of two cities’ in start-up funding 2025 year to date. Crunchbase data reveals that 11 mega-startups raising billion-dollar rounds have absorbed more than 33 % of all startup funding in H1 2025, most of which are AI focused, while 16 companies with $500M+ rounds claimed nearly a third of Q2’s capital. “Investor Ethan Kurzweil, a co-founder and managing partner at VC firm Chemistry, reacts to the data. Firms are anointing select startups as the consensus picks, then pushing them into bigger early-stage funding rounds to get their allocation of shares, he says. And while VCs might pride themselves on contrarian viewpoints, what they might consider to be such isn’t so differentiated anymore.”

FalkorDB included in Oxx rising stars of open source 2025 list.

Motorica features in Animation World Network and Cartoon Brew following funding announcement.

Aquant
Senior Enterprise Account Executive (New York)

Groundcover
Product Designer (Tel Aviv)

Jurnee
Software Engineer Intern (Paris or remote in France)

Sourcix
Lead Product Manager (Tel Aviv)