The Lilliputians Have AI Now: On SaaS and the Era of Disposable Software

• • The Lilliputians in Gulliver’s TravelsThe Lilliputians Have AI Now. On SaaS and the Era of Disposable SoftwareThis was a very interesting week where the long-held assumption that SaaS was the goose that lays the golden eggs was suddenly called into question. We saw stocks drop after Claude announced new plugins for Claude Co-work. It was a doozy of a week for announcements (the market rallied yesterday, adding to the craziness). There was OpenAI’s new Codex app and Codex 5.3. Anthropic launched Claude Opus 4.6 with agent-team orchestration and is rumored to be launching Sonnet 5 imminently. Anthropic even launched some awesome Super Bowl commercials (https://www.youtube.com/watch?v=kQRu7DdTTVA) poking fun at certain competitors. If this dizzying pace is indicative of where things are headed, there will be quite a lot soul-searching for software companies. And of course, there’s the SaaS slump in the markets, leading to fears of broader “contagion” (https://www.wsj.com/finance/investing/the-software-rout-is-spreading-pain-to-the-debt-markets-d6dd1397?st=8ddo34&reflink=desktopwebshare_permalink) - debt loads, high valuations, and investing in the right place at the wrong time. As always, our industry is about breakneck speed, killing our darlings, and thinking you’re being disruptive…until you’re disrupted. Software Isn’t Dead. But It Is Becoming Disposable.Is SaaS, and software in general, dead? It certainly feels like the vibe pendulum has swung this way. There is no doubt that the current crop of AI tools is progressing at warp speed, and the capabilities are improving all the time. If your context for AI is the tools from a few years ago - or even six months ago - you’re basically missing several lifetimes’ worth of changes. At the same time, I also feel there’s a bit of an overcorrection on the pendulum swing, which happens in times of extreme fear or exuberance. It’s undeniable that these AI tools are fundamentally changing how we consume and create software. As I wrote about in “Eroding the Edges (https://joereis.substack.com/p/eroding-the-edges-ai-generated-build),” we are moving toward point solutions for specific tasks. It’s not necessarily that I need a massive, complete SaaS product for everything anymore; I certainly don’t need their sales team breathing down my neck every renewal cycle. I just need to solve the problem at hand. In this new world, software becomes almost disposable. I can create whatever I want, deploy it, and tear it down. It doesn’t matter because at the drop of a hat, I can recreate most of the apps I built in a matter of minutes. It reminds me of the teardown disposability of containerized apps back in the day. I see this in my own work every day. I’m at the point where I have subscriptions to every major AI out there: Cursor, Claude, Gemini, ChatGPT, Descript, you name it. I’ll gladly pay for whatever tooling adds value. I was showing a friend the other day how to use Claude Code to construct an app he had been thinking about. By the time our ten-minute conversation was finished, the app was done. Was it production-ready? No. But it showed his idea was possible. With another couple of minutes, we got the app to solve his particular use case, and that’s the point. We are moving from an era of renting generic capabilities to generating specific ones. The assumption that SaaS was the golden goose was based on the idea that software was hard to build and easy to rent. That dynamic has flipped. Software is now easy to build, and increasingly hard to justify renting. The Jevons Paradox of Software and the Ten Minute AppBut here is the catch, and it’s a big one: for everything I automate, I end up creating ten more things on my plate. As I discussed in “Parkinson’s Law and AI (https://joereis.substack.com/p/parkinsons-law-and-ai-does-ai-meanmore),” we are in the era of abundance. I can automate all the annoying things in my life. But I’m finding that the annoying things don’t disappear. There are always more problems to solve. This is essentially Jevons Paradox (https://en.wikipedia.org/wiki/Jevons_paradox) applied to software. In the 1860s, the economist William Stanley Jevons observed that as coal engines became more efficient, total coal consumption didn’t decrease. Instead, it skyrocketed because efficiency made new use cases economically viable. The same thing is happening with software right now. As it becomes trivially easy to build, we don’t build less. We build dramatically more. Every problem that was previously “not worth writing software for” suddenly is. This creates an interesting tension with the “SaaS is dead” narrative. If the total surface area of software explodes, someone (or something) still has to manage, maintain, orchestrate, and secure it all. The ten-minute app I built for my friend works great. At least until it needs to talk to three other systems, handle edge cases at scale, or survive an audit. Disposable software is powerful for solving discrete problems. It’s less clear that it replaces the connective tissue that holds complex organizations together. That said, I expect AI agents to become increasingly capable at orchestrating this connective tissue and solving and coordinating enterprise-wide problems. Not now, but someday soon. Perhaps my new trillion dollar idea, the Vibe Stack (https://joereis.substack.com/p/the-vibes-stack-a-technical-deep), is the path to this future? Which brings me to where the “SaaS is dead” thesis breaks down. Where This Doesn’t ApplyNot all software value lives in the code. For many SaaS companies, the moat was never about the features. The moat was the data gravity, network effects, the integration ecosystem, and the operational reliability at scale. Nobody is going to vibe-code a replacement for Snowflake’s data sharing network or Salesforce’s org-wide workflow lock-in over a weekend. Switching costs are too high for most customers anyway. Plus, most organizations have a sprawling legacy of systems and code that will take humans several lifetimes to reconcile and update. I quipped last night on LinkedIn that a sign AGI is here is if it can migrate undocumented spaghetti codebases in mega-enterprises from legacy systems to modern ones. The vendors most exposed right now are the ones whose entire value proposition is a workflow wrapper, the ones that are essentially a UI on top of a database with some business logic sprinkled in. If I can describe what your product does in two sentences, AI can probably build it. If your product’s value is emergent from years of accumulated data, deeply embedded integrations, and institutional muscle memory, you’ve got more runway than the market is pricing in right now. That said, runway on its own isn’t guaranteed safety. I remember sitting in a conference hall this time last fall, looking at all the booths and chatting with a friend, when they said that 50% of these vendors won’t be around in the next few years. I still stand by that. In fact, the timeline might be even shorter. The Vendor Wake-Up CallIf you are a vendor, this is a massive wake-up call. The idea that you can build a company around a “feature” is being tested in real time. If I can have AI create that feature for me in five minutes, you don’t have a company. It reminds me of Gulliver’s Travels, where Gulliver is tied down by hordes of tiny people called Lilliputians. While not dangerous on their own, in a mass, they’re a force to be feared. Vendors are facing a potential Lilliputian attack of millions of people using AI to chip away at countless vendor flagship offerings from millions of tiny angles. We are seeing a change in how vendors go to market. Everyone has an AI agent now; everyone is using the same tools. Everyone is moving faster, but roughly at the same speeds. Net/net, nothing fundamentally changes in terms of competitive jockeying. Momentum is part of the equation, but you need a competitive advantage that is difficult to replicate, which is getting harder and harder to achieve. This extends to the “legacy” question as well. The future won’t be evenly distributed. The slow rate of change at big, legacy, and enterprise companies can, in some cases, serve as a moat. Sloth and inertia are powerful forces, especially if the company dominates a vertical. But this also means these companies could disappear quickly, as history has repeatedly and violently shown. Creative destruction and stuff. There’s also the regulatory angle. Even in highly regulated industries - healthcare, finance, defense - I think things will move faster than people expect or wish. Not because regulation will disappear, though it might (modern administrations are quite anti-regulation right now). The cost of failure in these industries is catastrophic, and liability frameworks exist for real reasons. But regulation has a history of lagging behind technological change by years, sometimes decades. And during those lag periods, a thousand small insurgents armed with AI can chip away at an incumbent’s flagship offerings from a million tiny angles. Think of how fintech slowly unbundled banks, except this time the cycle is compressed from a decade to a couple of years. If you’re an incumbent relying solely on regulation to keep you safe, dream on. And because you’re sleepwalking, you won’t even know until it’s too late. What This MeansFor engineers and problem-solvers, this is the era of abundance we’ve been waiting for. The friction between “having an idea” and “solving the problem” has never been lower. But for the software industry as a business model? The easy money is gone. The vendors that survive won’t be the ones with the most features or the best placement on a legacy research analyst quadrant. They will be the ones tackling the problems that are still too complex, too deeply embedded, or too data-intensive for me to solve with a ten-minute conversation with a chatbot. And for the rest of us - the builders, the tinkerers, the people who see a problem and just want to fix it, the real question isn’t whether SaaS is dead. It’s whether we can keep up with our own appetite for creation. Because the tools have caught up with our imaginations, and it turns out our imaginations are insatiable. Software isn’t dead. But the “Software as a Service” free ride? That’s definitely over. Also, listen to this as a podcast. Available on Spotify, Apple, and wherever else you get your podcasts. Please support the show with a review. It means a lot. The Practical Data Community 2026 State of Data Engineering Survey results will be released next week. This is an independent and no-BS look at the reality facing practitioners today. No vendors or clueless pay-to-play analyst research firms influenced this survey.

Lots of very interesting and surprising gems in there. There might even be a hackathon…stay tuned. • • Also, if you’re a vendor or event looking to work with me (product reviews, sponsorships, talks, etc), message me. I’ve put together some brand new offerings that you might be interested in. Have a great weekend, Joe Thanks for reading! Subscribe to receive new content, usually every week. Awesome Upcoming EventsConfluent (https://www.confluent.io/). San Francisco. Me. March 26th…we’ve got something special in store for data engineers. Stay tuned for an announcement. Some friends of mine are doing these events: SLC Kyle Nesbit (CEO of Credible (https://credibledata.com/), ex-Google) will talk about Giving Data Value to AI at the Utah MLOps Meetup on Tuesday, February 24th. Register here (https://www.meetup.com/machine-learning-utah/events/311226939/?eventOrigin=network_page). Mountain View My good friend Demetrios and the ML Ops community are doing an amazing event on Tuesday, March 3rd in Mountain View, at the Computer History Museum. Coding Agents: AI Driven Dev Conference. Register here (https://luma.com/codingagents). As for me…Still working on my 2026 event schedule, and so far it looks dope. Will reveal more soon, so stay tuned… But wait, there’s more!• • Cool Videos and ReadsAnalytics agents are all the rage. And so are semantic layers. Cube (https://cube.dev/) and I partnered on my review of their new analytics agent. It was a lot of fun to try to beat up their AI agent, which did quite well in my tests. Mike Driscoll (Founder and CEO of Rill Data) and I had a great chat about the state of AI BI, the resurgence of semantic layers, and much much more. I was recently a guest on AI with Arun. We discussed Franken-stacks, building useful things, and much more. Fun times. Here are some things I read this week that you might enjoy.How OpenClaw’s Creator Uses AI to Run His Life in 40 Minutes | Peter Steinberger (https://www.youtube.com/watch?v=AcwK1Uuwc0U) Leaders, gainers and unexpected winners in the Enterprise AI arms race (https://www.a16z.news/p/leaders-gainers-and-unexpected-winners) Did A.I. Take Your Job? Or Was Your Employer ‘A.I.-Washing’? - The New York Times (https://www.nytimes.com/2026/02/01/business/layoffs-ai-washing.html?unlocked_article_code=1.JFA.8k3A.GT2R5S9B-zex&smid=url-share) Selfish AI | GarfieldTech (https://www.garfieldtech.com/blog/selfish-ai) New Data: OpenAI’s Lead Is Contracting as AI Competition Intensifies (https://www.bigtechnology.com/p/new-data-openais-lead-is-contracting) Floe and Apache Polaris: Policy-Driven Table Maintenance for Apache Iceberg (https://polaris.apache.org/blog/2026/02/04/floe-and-apache-polaris-policy-driven-table-maintenance-for-apache-iceberg/) What does the disappearance of a $100bn deal mean for the AI economy? | AI (artificial intelligence) | The Guardian (https://www.theguardian.com/technology/2026/feb/05/disapperance-100bn-deal-ai-circular-economy-funding-nvidia-openai) Claude Code is the Inflection Point (https://newsletter.semianalysis.com/p/claude-code-is-the-inflection-point) SaaSmageddon and the Super Bowl – Stratechery by Ben Thompson (https://stratechery.com/2026/saasmageddon-and-the-super-bowl/) Find My Other Content Here📺 YouTube (https://www.youtube.com/@joereisdata) - Interviews, tutorials, product reviews, rants, and more. 🎙️ Podcasts (https://open.spotify.com/show/3mcKitYGS4VMG2eHd2PfDN?si=7bd8ffc4d5b840ce) - Listen on Spotify or wherever you get your podcasts 📝 Practical Data Modeling (https://practicaldatamodeling.substack.com/) - This is where I’m writing my upcoming book, Mixed Model Arts, mostly in public. Free and paid content. The Practical Data CommunityThe Practical Data Community is a place for candid, vendor-free conversations about all things tech, data, and AI. We host regular events such as book clubs, lunch-and-learns, Data Therapy, and more. 🤖 Join on Discord (https://discord.gg/gNfw5AKWSK)

Thanks for reading! Subscribe for free to receive new posts and support my work.