The Twenty Minute VC"Cursor is Dead" is Total BS: Here is Why | Miles Clements
CHAPTERS
How to find real alpha in AI: time-to-value vs durability
Miles lays out a simple framework for evaluating AI companies: how fast users get value, and how durable that value remains once adopted. He contrasts slower-deploying vertical AI (legal/accounting) with “vibe coding” apps that deliver instant results but often lack staying power. Coding tools stand out because they score highly on both dimensions.
- •Framework: time-to-value and durability of value
- •Vertical AI can be slow to deploy but highly sticky once embedded
- •Vibe coding is fast but often lacks durable differentiation
- •Coding is the current AI battleground because it compounds value quickly and durably
Why “Cursor is dead” is narrative, not reality
Harry presses on claims that developers are abandoning Cursor for Claude Code and that pricing will force churn. Miles argues the category is expanding so quickly that competitors can grow simultaneously rather than purely stealing share. He also notes that much of reported ARR growth is consumption-driven, not just seat-based expansion.
- •Claude Code’s momentum is real but doesn’t automatically imply Cursor decline
- •The market is expanding: more new developers and more usage per developer
- •ARR growth is heavily driven by consumption, not only new seats
- •Social-media narratives can lag actual usage and revenue dynamics
Cursor’s shift from IDE novelty to agent-first workflow
Miles says Cursor is misunderstood as merely an IDE with autocomplete; in reality, the product is increasingly agent-centric. He cites internal/posted metrics to argue agents are now the primary usage mode and are rapidly scaling in real production workflows. The team’s focus and execution matter more than external commentary.
- •Cursor is “a victim of its own success” as people anchor on the IDE feature set
- •The industry is moving from copilots to agents; Cursor has been explicit about this
- •Cited metrics: more agent usage than tab completion; high daily agent engagement
- •Cloud agents increasingly contribute to merged PRs and production output
Cost, model dependence, and the case for multi-model products
Harry challenges whether Cursor’s reliance on Anthropic models creates untenable costs. Miles argues Cursor’s multi-model approach is a structural advantage because developers switch models frequently and want flexibility. Being an “index” of model innovation creates a compounding flywheel: product improvements plus model improvements both raise capability.
- •Multi-model is valuable because developers switch model families frequently
- •Cursor benefits from model progress across providers, not just one lab
- •Compounding flywheel: Cursor feature upgrades + underlying model upgrades
- •Cost concerns exist, but flexibility and user preference support multi-model strategy
Why building Cursor’s own models can still be rational
Miles defends Cursor’s effort to build proprietary models, framing them as specialized coding models rather than general-purpose LLMs. The goal is differentiation on professional, enterprise-grade coding tasks, not broad intelligence benchmarks. Specialized models can improve quality, control, and enterprise suitability.
- •Generalist vs specialist framing: Cursor aims for coding specialists
- •Enterprise needs task-optimized models more than general capabilities
- •Differentiation: models that improve professional workflows, not “poetry”
- •Owning part of the stack can deepen product defensibility over time
Underwriting Cursor at high prices: platform ownership and “engineering OS” upside
Miles explains the upside case for investing at very large valuations: Cursor could become the first true platform company for engineering, owning a broad vertical the way Salesforce owned go-to-market. He argues prior winners (Atlassian, Datadog) captured slices of the stack; Cursor’s ambition is platform breadth. He also notes that extreme revenue growth can make traditional multiple debates feel less central.
- •Thesis: first platform company for the engineering vertical
- •Historical analogy: slice winners (Atlassian/Datadog) became $50B–$100B outcomes
- •Focus on product-market fit signals; financials reflect intensity of pull
- •Traditional valuation rules feel strained when growth is unprecedented
Forecasting in hypergrowth: assumptions matter more than the plan
After discussing how far off early ARR forecasts were, Miles reframes planning as a way to encode assumptions rather than to police quarterly targets. He contrasts venture investing with public-market earnings management. The key is understanding inputs—product, pricing, segment penetration—more than variance to budget.
- •Revenue plans matter as assumption maps, not as enforcement tools
- •Venture investors shouldn’t manage like public-market investors
- •Focus on inputs (product/pricing/segment mechanics) vs output precision
- •Hypergrowth makes “plan vs actual” less informative than it used to be
Do “non-15x” companies still matter? Consensus vs non-consensus and portfolio nuance
Harry argues big funds require huge exits, making slower growers unattractive; Miles pushes back that outcomes depend on many inputs beyond growth rate. He warns investors get “hammered in the middle” and argues you can win via either high-consensus leaders or differentiated, non-consensus bets with better ownership. The key is embracing nuance across stages and strategies.
- •Growth rate alone ignores founder quality, market, valuation, and ownership
- •Polarization: AI maximalism vs sitting out; Miles advocates nuance
- •Multi-stage, multi-strategy approach can combine leaders + overlooked opportunities
- •Portfolio construction matters: extremes can work; the middle is dangerous
Painful misses and the lesson of ServiceTitan and Rippling
Miles recounts losing ServiceTitan due to rigid pricing/multiple rules, arguing that deep market understanding should sometimes override heuristics. He then discusses missing Rippling, focusing on “marginal ease of ARR accumulation” and Parker Conrad’s ability to build compounding growth mechanisms through adjacent products. He attributes the miss to reputation drag, speed, ownership thresholds, and rule rigidity.
- •ServiceTitan: over-indexing on multiples led to missing a major outcome
- •Rippling: compounding growth physics and adjacent revenue lines drove durability
- •Founder reputation and fast-moving rounds can slow decision-making
- •Ownership rules can be costly when an opportunity warrants rule-breaking
Coverage, win rates, and how Accel self-corrects after missing breakouts
Miles describes how Accel runs rigorous partnership reviews: identifying top private companies, scoring “investor of record” positions, and diagnosing misses. He estimates an 80% term-sheet win rate as healthy and argues losing sometimes signals you’re competing for the right deals. The discussion also covers the shrinking relevance of old benchmarks and the importance of usage intensity over headline growth.
- •Global partnership reviews: “How did we miss it, and how do we fix it?”
- •Goal: high coverage and high win rate, even if perfection is impossible
- •Healthy win rate is not 100%; losing can indicate competitiveness
- •Benchmarks are obsolete; usage intensity and engagement matter more than ever
Anthropic: underwriting mega-outcomes and the Pentagon controversy
Miles explains that a small set of private companies operate on a different plane where trillion-dollar outcomes are plausible, making late-stage participation rational. On Anthropic’s Pentagon-related controversy, he avoids specifics but praises founders for sticking to principles and mission under commercial pressure. They discuss how values-driven decisions can test loyalty, talent, and adoption dynamics.
- •A few companies (labs/defense leaders) can credibly be trillion-dollar candidates
- •Investing is not about “squinting to 3x” for these rare platforms
- •Founders’ principles get tested when major contracts appear
- •Public response can be complex; mission alignment can still build trust
What happens to 2021-priced companies: LBO land, tough outcomes, and founder-led advantage
Harry asks what happens to companies like Miro and Snyk that must live up to high private marks amid slower growth and repriced public comps. Miles argues many are still great businesses but may find different “homes,” often via PE buyers such as Thoma Bravo and Vista. They discuss why founder-led companies often inspire higher conviction, while also acknowledging exceptional professional CEOs can succeed.
- •2021 valuation marks created difficult expectations, even for solid businesses
- •PE/LBO firms may be the natural buyers for many reset-name outcomes
- •Founder behavior under pressure is crucial for navigating downcycles
- •Founder-led tends to be special, but elite professional CEOs can outperform too
Going public now: liquidity, currency, and the ‘sub-$5B’ trap
Miles explains why founders stay private longer: secondaries can provide employee liquidity, and M&A currency/valuation signaling can also exist privately—especially for top-tier companies. He argues the hardest zone is IPO’ing too small (roughly sub-$5B), where many companies struggle to break out in public markets. The decision to go public is increasingly about scale and clarity of growth beyond that threshold.
- •Private markets now offer liquidity and currency features once unique to IPOs
- •Only the best private companies can fully replicate public-market benefits
- •Companies IPO’ing around ~$2–$5B often struggle to “break out”
- •Founders wait for clearer line of sight to trading above key scale thresholds
Liquidity decisions, evergreen funds, and how VCs should handle public positions
They debate when to take chips off the table, with Harry citing cases where selling early was rational despite company preferences. Miles says it’s situational: tenders can be smart for diversification, but long-duration compounders like CrowdStrike rewarded continued ownership through multiple stages and the IPO. On evergreen/public positions, he argues multi-stage firms should stay with outlier founders, but not every company compounds long-term.
- •Principle: prioritize what’s best for the company; diversify when appropriate
- •Examples show both sides: WeWork-style risk vs CrowdStrike-style compounding
- •Multi-stage investing enables laddering into meaningful long-term ownership
- •Evergreen/public holding can work selectively; not as a blanket rule
Inside Accel’s craft: sourcing, board dynamics, and career advice
The conversation closes with practical insights: who Miles thinks are Accel’s best at sourcing, picking, and winning deals, and what makes a great board member. He emphasizes that founders don’t need micromanagement, but do benefit from a trusted sounding board for a few major “bumper decisions” each year. He ends with advice on professionalism and the importance of consistent firm rituals over time.
- •Roles: best sourcer (Christine Esserman), best picker (Andrew Braccia), best closer (Samir Gandhi)
- •Board value is highest on a few big decisions per year, not day-to-day tactics
- •Great board members deliver feedback with humility; loud voices often add less value
- •Career advice: sustained professionalism and respect for process are enduring edges
