AINEWS
Drag and drop app builders like Airtable and Notion are being absorbed by AI agents that read the underlying data and decide what to do, an ex Airtable engineer argues. The data model survives; the automation layer doesn't.
Bending Spoons closed its $1.28 billion acquisition of Airtable this week, a sharp discount to the $11 billion valuation Airtable commanded in 2021, per TechCrunch and Inc.. The deal is the first hard pricing signal that declarative-schema no-code, the category of drag-and-drop app builders like Airtable and Notion, is being absorbed into LLM tool-use loops. The user-defined data model survives. The point-and-click automation layer on top of it doesn't.
An engineer who spent years inside Airtable argues the $1.28B price tag is what happens when that category gets absorbed by an LLM tool-use loop. His post on the exe.dev blog is opinionated and worth reading with the author's current employer in view. He now works at exe.dev, which sells an AI sandbox aimed at roughly the same buyer Airtable once sold to. The post is also a single informed view, not consensus. The HN thread is bifurcated. Some readers called it a "puff piece" with 20% content and 80% pitch; others called it a clearly stated indicator of a massive shift. The disagreement is about the size of the displacement, not its direction.
What is the mechanism? A user-defined data layer plus a flexible wiring layer on top. No-code platforms won their initial market by letting a non-engineer declare a schema (the tables and columns) and then layer rules, views, and automations over that schema. The schema is durable. It encodes the business model in a way the author quotes Fred Brooks as capturing decades ago: "Show me your tables, and I won't usually need your flowcharts." The wiring layer is what an LLM is now reaching for. When the model can read the schema and decide how to update, route, or notify, the point-and-click automation surface stops being a load-bearing product.
The historical arc backs the durability half. FileMaker launched in 1985 on the same idea: a non-programmer describes tables, the platform does the rest. Salesforce's 2000 "Software is Dead" campaign made the same bet about replacing bespoke IT with declarative SaaS. Each time, the user-defined data layer outlived the surface that sold it. The exe.dev post reads the present moment as another instance of the same pattern, with an agentic interface plus human review plus light editing replacing human-centric point-and-click as the dominant surface.
For the employee who used to be the one who brought Airtable to the team, the new stack is actually more general. They can describe a process in plain English, point a model at a Notion table, and ship a working automation without waiting for IT. That is agency. The trade-off the source does not wave away is that the LLM-driven version is non-deterministic, harder to audit, and introduces a new kind of lock-in around whichever model and tool ecosystem the team adopts. The IT-as-procurement role exists for exactly those reasons, and the structural argument is not that IT disappears but that its center of gravity moves from buying seats to governing model behavior, data access, and reproducibility.
The falsifier is concrete. If Airtable's installed base quietly renews at scale under Bending Spoons with no migration pressure to LLM-driven stacks, the category-pricing verdict weakens. The data-model-outlasts-UI half of the claim still holds either way. The interesting question for 2026 is which side of that split the next Airtable renewal cycle lands on, and whether Bending Spoons' integration roadmap reads to customers as a continuation of the Airtable product or as a managed decline.