1.

Doug Klunder's recalculation algorithm turned a spreadsheet into a reactive instrument so intuitive that 750 million people became programmers without knowing it. The result: every B2B company you admire is either unbundling Excel or trying to bottle its soul.

2.

Two innovations won the Spreadsheet Wars

• Klunder's two-dimensional simultaneous recalculation meant only affected cells recomputed — a devastating speed advantage over Lotus 1-2-3's full-sheet recalc, and the origin of Excel's signature magic: change an input, watch everything respond
• Gates and Raikes switched mid-project from PC command line to Mac's graphical interface, bringing spreadsheets "from PhD thesis mode into something that an average person could use" (Jon Devaan)
• Lotus did $53 million in year one and tripled to $156 million — then lost to Excel, which launched exclusively on Macintosh in 1985 and led by revenue in 1991

3.

750 million unwitting programmers

• Excel is declarative: type =IRR(C4:G4) without knowing the formula, while Microsoft's developer army optimizes the engine underneath. Each update improves the how without the user lifting a finger
• It is reactive: change one cell and every dependent output recalculates instantly. Kahneman's condition for skill acquisition — "rapid and unequivocal feedback" — is baked into the product's architecture
• A single workbook is naturally full stack — one sheet as database, another as transformation logic, a third as user interface with charts and controls — which is why a lone investment banking analyst can build an end-to-end financial model in an 80-hour week without outside help

4.

Quality and network effects compound for 36 years

• Excel holds ~80%+ market share with a near monopoly for intensive use cases — "send them your model in Google Sheets, and see how seriously they take your idea"
• Two-sided platform lock-in: banks train analyst classes on Excel, those analysts start companies mandating Excel, plug-in developers build for Excel, and .xlsx is the universal format across every investment bank and hedge fund on earth
• The Lindy Effect predicts software surviving 36 years should survive 36 more — and Excel's case is not mere persistence but earned dominance, refreshed generation after generation

5.

Flexibility without guardrails creates opportunity

• No domain awareness: botch a FIFO inventory model and you blame yourself, not the software — purpose-built tools prevent those errors by design, which is exactly the gap entrepreneurs exploit
• No provenance: copy-paste into a database tab destroys the chain of computation, making reproducibility impossible — every missing audit trail is a SaaS company waiting to be born
• LAMBDA made Excel Turing complete on February 9, 2021, but as the authors concede, it "sharpens the double edged sword" — more complex programs, same versioning and readability problems, more entrepreneurial kindling

6.

Unbundling built $500 billion in market cap

• Tomasz Tunguz (Redpoint, 2017) and Ross Simmonds (2019) mapped the phenomenon: Salesforce at $193 billion, plus Asana, Tableau, Workday — nearly half a trillion dollars from companies replacing specific spreadsheet workflows
• The mechanism: entrepreneurs watch what 750 million users hack together in spreadsheets, size the market, build purpose-built tools — "practically any software that is built to handle data that isn't super long text is unbundling Excel"
• Yet Excel's core — financial modeling, ad hoc analysis, rapid prototyping — remains untouched, its market share unmoved, its user base undiluted

7.

"Inspired by Excel" captures the design philosophy

• Rich Hickey (creator of Clojure) argues good software should do one thing well, like a saxophone — "Nobody wants to play a choose-a-phone." Excel is the choose-a-phone that hundreds of millions compose for
• Airtable ($2.5 billion valuation, September 2020) became a platform for solutions its founders never imagined; Figma's "WFH Town" proved a design tool could become a virtual world; Bubble lets users create production apps by linking a spreadsheet
• Microsoft paid a coincidentally identical $2.5 billion for Minecraft in 2014 — analysts were confused, but the logic is the same: flexible, user-driven creation tools have staying power that purpose-built products cannot replicate

8.

Excel's catastrophic failures prove fragility, not influence

• The London Whale: JPMorgan lost $6 billion partly because of a spreadsheet error in a Value-at-Risk model. Reinhart and Rogoff's austerity paper, which shaped fiscal policy across Europe, contained an Excel coding mistake that excluded key countries from the dataset
• The UK's Public Health England lost nearly 16,000 COVID case records in September 2020 because Excel's row limit truncated the data — not a user error but a product constraint in software the essay calls "the most influential ever built"
• An essay arguing Excel's influence is underappreciated that omits its influence on catastrophic failures is not incomplete — it is self-refuting. The same flexibility the authors celebrate is the mechanism that produces civilizational-scale errors, and no amount of Lindy resilience changes the fact that software without guardrails becomes more dangerous, not less, as more of the world depends on it

9.

Dangerous infrastructure is still infrastructure — that's the point

• Both the thesis and the counter-argument assume influence and safety travel together: the authors celebrate Excel's reach, the critics weaponize it. But the most influential technologies in history — automobiles, antibiotics, nuclear energy — are precisely the ones whose failure modes are catastrophic
• No one argues cars should have remained niche because 1.35 million people die in traffic annually. The death toll is evidence of ubiquity, not an argument against it. Excel's errors at JPMorgan and Public Health England are legible only because Excel is load-bearing infrastructure for civilization
• The real question is not whether Excel is dangerous — it is — but whether any replacement could achieve its combinatorial reach without reproducing its risks. Forty years of purpose-built software suggest the answer is no: every tool that adds guardrails subtracts flexibility, and the flexibility is what makes 750 million people show up