The Joy of Tech: A Different Path for Products, People, & AI | by Dave Feldman | Jun, 2026 | MediumSitemapOpen in appSign up<br>Sign in

Medium Logo

Get app<br>Write

Search

Sign up<br>Sign in

The Joy of Tech: A Different Path for Products, People, & AI

Dave Feldman

15 min read·<br>19 hours ago

Listen

Share

Sometimes the tech industry feels like one bubble after another: dot-com, mobile-social-local, chatbots, crypto, web3, prediction markets…but the AI bubble is surely the weirdest yet.<br>Weird because the excitement is callous and self-contradictory: “This technology will destroy the world — so we’re building it as quickly as we can!” — ”We’re just so thrilled to launch this thing that will take away all your jobs!” There’s a lot of success riding on a strange substrate of insensitivity. I won’t claim to know where it’s all going. But I do work with AI every day, and I know enough to set the record straight about a few things.<br>Yes, AI can build prototypes faster. Yes, rigid design process is bad and AI can help. Yes, AI is good at writing code. And no, AI can’t replace PMs or designers or engineers.<br>We can’t just get rid of these professions, because doing so will result in bad products — for every definition of “bad” you can think of.<br>Except…maybe we don’t care if our products are any good. Maybe we don’t care about our products at all.<br>Where’s the product?<br>It’s been a while since we did care. This isn’t new to the AI bubble, though it’s become a lot more evident: amidst all the bragging — about how we’re building (with AI), and who’s building (not you, you’ve been laid off…it’s AI again) — there’s virtually no mention of what we’re building ❶.<br>Does that seem strange to you? If we’re vibe-coding new platforms overnight, if we’re building at 10x speed and have been for the past year, then we should have years’ worth of amazing products at our fingertips. Where are they?<br>Many of the new products I do see are really bad. Forget product-market fit: they simply don’t work, in the basic sense that you click on stuff and nothing happens.<br>According to just-published research from MIT, mobile-app releases have increased in the era of “Agentic AI” but usage and satisfaction have dropped:<br>Press enter or click to view image in full size

Meanwhile, preexisting products are deteriorating: GitHub has become unreliable enough that it’s a joke; the comment field broke on my LinkedIn for a week; Claude is erroring out on 20% of my prompts; TikTok wouldn’t let me dismiss the cookie popup so I simply couldn’t use it; My iOS updates were silently failing for a month; my entire computer has started freezing periodically ❷.<br>Some of you are going to tell me it’s my fault. I’m holding it wrong. There can’t possibly be all these bugs because your stuff is working just fine ❸. But I’ve been asking around, and others are noticing the same thing. The lights are going out all over the internet, and the people who are supposed to care…don’t.<br>I’m not saying we’re not building stuff. We’re building a lot of stuff ❹, the majority of which is neither well-built nor particularly useful.<br>We brag about how fast we’re moving and my response is, “Yes. It shows.”<br>Layers of indirection<br>How did we get here? Maybe the history matters less than where we’ve landed — and in particular, how our philosophies make it reasonable not to care about the product. Or the customer. Or our teams. Think about it as a set of layers:<br>Layer Zero: Product Development<br>Like a lot of folks in tech I’m motivated by the fun of building software, the challenge of creating useful, high-quality products, and the joy of doing that with other people. One can make a living at it because when it’s done well, it produces something people are happy to pay for.<br>Product development involves a lot of uncertainty. Answering even its most basic questions requires we balance a myriad of factors — some numeric, some quantitative, some neither — and make trade-offs partially rooted in subjective judgment and expertise. We have a lot of frameworks to help us, so it can be a manageable, structured process. But for all the ink spilled on articles with titles like, “The Foolproof Engine for Finding Product-Market Fit,” there is no recipe for how to do it. That’s OK; it might even make it more fun.<br>Layer One: Mechanization<br>Peter Drucker famously said, “If you can’t measure it, you can’t manage it,” and the tech industry has dialed that up to eleven. Our goals must be measurable, our objectives broken down into key results, our productivity quantified. It’s a comforting approach for computer nerds like us, drawn to the precision of the digital and often alienated by the squishy human world around us.<br>So we demand numbers over narrative, quantitative over qualitative…and critical information is lost. Our MAUs, engagement, retention, and NPS scores do not equal our product-market fit or product quality — not on their own. But boy can we put them on a graph.<br>A mechanized tech company is one that wants...