AI was supposed to change everything — so why are we still using the same apps? A question posted on Hacker News ignited a firestorm in the developer community: "Where are all the disruptive software that AI promised?" Hundreds of comments poured in, and the responses were surprisingly honest.
What Is This About?
HN user p-o asked a simple question: "If AI is such a game changer, where are the apps? I'm still using the same stuff as I did 5 years ago." Synthesizing the responses from developers, founders, and PMs, a structural gap between AI expectations and reality becomes clear.
The core arguments split into three camps.
First, the view that only code production costs dropped. HN user veunes nailed it: "AI only drove down the cost of writing code, not the cost of finding Product-Market Fit". You can spin up a Notion clone over the weekend with Cursor, but getting users to migrate their data, change their habits, and pay for it is just as brutally hard as it was a decade ago.
Second, the UX limitations of probabilistic systems. Another user put it this way: "When I click Save, I know exactly what's going to happen. When I type a prompt into an AI agent, I'm basically playing roulette every single time". For mainstream users accustomed to deterministic UX, probabilistic AI still feels more like a toy than a tool.
Third, the view that change is happening — just invisibly. An Italian arborist claimed he completely transformed his company with AI in just 4 months, and others pointed to internal apps and CS tools as the real frontier of quiet innovation.
What's Actually Different?
So what has actually been disrupted by AI, and what hasn't? Here's a research-backed breakdown.
| Area | Where AI Change Is Real | Where Expectations Fall Short |
|---|---|---|
| Coding Assistance | Copilot/Cursor boosted repetitive coding productivity 2-3x | Complex architecture design and large-scale system ops remain human territory |
| Translation | Real-time translation quality described as "telephone-level" breakthrough | Cultural nuance and specialized legal/medical translation still require human review |
| Internal Automation | Billing, QA, document summarization spreading fast | Meaningful differentiation in customer-facing products remains rare |
| SaaS Market | Existing apps adding AI features (search, recommendations, auto-classification) | "AI-native" killer apps replacing existing SaaS have yet to emerge |
| Agentic AI | Rapid progress at pilot/experiment level | Only 10% of organizations have successfully scaled agents to production |
| Games/Creative | Explosion of image and music generation tools | Not a single hit indie game made with AI has emerged |
MIT Sloan Management Review's Davenport and Bean explain this through Amara's Law: "We tend to overestimate the effect of a technology in the short run and underestimate the effect in the long run". GenAI has already entered Gartner's Trough of Disillusionment, and agentic AI is predicted to follow the same path in 2026.
BCG's analysis is more specific: within 2-3 years, 50-55% of US jobs will be "reshaped" by AI — but this means transformation, not replacement. Parts of existing work get automated while humans perform the rest at a higher level.
According to NC Tech's analysis, BCG's 10-20-70 framework provides the answer: only 10% of AI value comes from algorithms, 20% from tech infrastructure, and the remaining 70% from people, processes, and organizational alignment. Deloitte's 2026 report also reveals that 84% of companies haven't redesigned work around AI. The technology is ready — the organizations aren't.
The Essentials: 3 Things to Do Now
Let's answer the "so what?" question with actionable strategies you can execute right now.
- Start with "invisible innovation"
Don't force AI into customer-facing products. Prove ROI first with internal workflow automation — billing, QA, document summarization. SaaS founders are reaching the same conclusion: AI doesn't need to be customer-visible to generate value. The most effective AI works quietly in the background. - Invest in PMF, not code
Being able to vibe-code an app over the weekend means nothing. Code is now a cheap commodity; distribution and trust have become exponentially more expensive. Focus on user validation speed over prototyping speed. In Designli's SaaS founder survey, the biggest concern was "building unvalidated AI features too early". - Make your organization AI-ready
Change the organization before changing the model. AI pilots never scale without work redesign. Cisco's AI Center Director warns: "When AI enters a process, it alters the process boundaries themselves". Retraining, governance, and change management matter more than algorithms.
Going Deeper
Will the AI bubble burst?
MIT SMR's Davenport and Bean see an AI bubble burst as "inevitable". The parallels to the dot-com bubble are clear — startup overvaluation, growth-over-profit emphasis, aggressive infrastructure buildout. But just as the fiber optic cables laid during the dot-com era became the backbone of the modern internet, AI infrastructure investments will likely create long-term value.
Will coding jobs disappear?
LSE researchers warn that "jumping from model capabilities to job replacement is a logical error". Just as VR promised workplace revolution but landed only in education and manufacturing niches, technological capability alone doesn't guarantee commercialization and adoption. What coders actually worry about isn't getting fired — it's becoming "AI babysitters" with diminished autonomy and expertise.
Is SaaS dead?
Per NC Tech's analysis, SaaS stocks have dropped ~30% from peak, but this signals market maturation rather than software's demise. The next shift will likely be from per-user pricing to workflow- and outcome-based pricing. Applications won't disappear — only how software is delivered, consumed, and monetized will change.
