All Software Will Be Rewritten

Roughly five years ago, the tech industry treated software moats as sacred. Complex codebases, proprietary interfaces, years of accumulated domain logic — these were considered unassailable defences. That assumption is now a joke.

A $1 trillion wipeout in software stocks wasn't panic — it was a fundamental realisation. Those moats were made of sand. And LLMs are the tide.

While public companies bore the brunt, thousands of private software firms backed by global VCs face an even steeper valuation cliff unless they pivot toward AI-native strategies. AI agents can now read, understand, and rewrite all software, making the traditional “SaaS moat” obsolete.

The Great Collapse of the Interface

The “Learned Interface” was historically the strongest psychological moat. Think of Bloomberg terminal power users who spent years mastering arcane codes — GP, FLDS, GIP — or COBOL programmers locked in by muscle memory rather than data.

On February 23, 2026, IBM shares tanked nearly 13.2%, wiping $40 billion in market value after Anthropic announced Claude Code could read, analyse, and rewrite legacy COBOL systems. What was once a high-cost, consultant-heavy fortress of expertise has been upended by AI's capacity to understand and rewrite that code.

LLMs have collapsed proprietary interfaces into a single universal entry point — natural language. Software that once required years of training now operates through conversation.

From Hardcoded Logic to Markdown Skills

This is the most devastating shift. Traditionally, vertical software encoded industry knowledge through thousands of hardcoded if/then branches — the crown jewels of IP. Now, a “Headless Software” model emerges:

Domain experts now describe workflows in plain English. Logic moves from compiled code to editable .md files or MCP-connected Notion pages. Iteration speed increases exponentially, and incumbents lose their historical speed advantage.

The New Architecture: MCP and The Agentic Bundle

Tasks that once required many seats and logins now resolve via a single natural-language command. Agents bypass UIs and orchestrate across silos, commoditising enterprise software's core promise into a conversational layer.

The silver lining: orchestrating this agentic future is extremely difficult. Curating MCP connectors, writing and validating thousands of markdown skills, governing sub-agent coordination, hardening memory against hallucinations, embedding audit trails, and integrating legacy data represents brutal, multi-year engineering work.

The SaaSocalypse destroys old castles — but the rebuild represents a vast frontier rewarding domain expertise and execution.

Code Evolves: From Static Artifact to Living Organism

The old model: code written, compiled, frozen until the next release. The new paradigm: code becomes dynamic, self-interpreting, and alive — a living prompt-and-memory system evolving with every interaction.

Humans become CEOs of agent swarms. The industry shifts from static logic to continuous interpretation and agent-owned relationships. Three implementation waves unfold:

Feb 2026 Market Reckoning — A Structural Repricing of SaaS Moats

The trillion-dollar event was a structural shift in how markets assess software defensibility. Value is now assessed on “rewritability” — where a company sits in the stack and what moat anchors it.

The Future — The “Headless” Enterprise

The rewrite spans the full legacy SaaS landscape: CRM (Salesforce, HubSpot, Zoho), ERP (SAP, NetSuite, Dynamics), HRMS (Workday, ADP), Accounting (QuickBooks, Xero, Tally), BI (Tableau, Power BI), Legal (DocuSign, Ironclad), Support (Zendesk, Intercom), and many more.

The barrier to entry has exploded — moving from a dozen dominant incumbents per vertical to hundreds of AI-native competitors. Winners will be determined not by code ownership but by unique data and deep customer trust.

Will you be the one holding the pen — or be rewritten out of existence?