Why Cursor Is Dead, An AI Tsunami Is Coming & You Need to Prepare

Summary

Why does it matter? Because the software industry is being repriced in real time — and most investors haven't noticed yet.

Jerry Murdock has co-founded and scaled a $90 billion firm over 30 years. When he says Cursor is already obsolete and that agents, not humans, will be the primary buyers of software, that's not a hot take — it's a dispatch from inside the earliest adopters. This episode maps what happens next: to pricing models, to systems of record, to private equity, to labor markets, and to Nvidia's long bet on ASICs.

• Autonomous agents have replaced human developers at several AI-native portfolio companies — and this transition is only about two months old
• Systems of record are not defensive moats; they are raw material that only execution can convert to durable value
• Per-seat SaaS pricing structurally breaks when the buyer is an agent optimizing on consumption
• White-collar displacement hits small and medium businesses first — and could decide the 2028 presidential election

Cursor is worth $27-30B — and the most advanced AI shops already consider it obsolete

Two-month-old autonomous agent workflows have lapped a multi-billion dollar company. That's Jerry Murdock's opening shot, delivered without hesitation. Companies in his portfolio — E2B, Eventual, Lotus AI, Oven, Get Dynasty — have spent the last two to six weeks replacing human developers with autonomous agents running on OpenClaw, NanoClaw, or homemade equivalents. "Most of the companies I just mentioned, their view, as they've told me, is Cursor is obsolete."

Murdock isn't writing Cursor off entirely — he thinks the team is smart, flush with cash, and has time to pivot toward autonomous agents. But the speed of the transition is the real story. These aren't enterprise giants running multi-year pilots; they're native AI startups that spun up agent-powered dev workflows before most investors had finished writing their AI thesis decks.

The implication cuts across every coding tool, IDE, and developer productivity product currently raising at premium multiples: the human developer is already not the primary customer at the frontier. If you're building for human buyers in this category, you're not six months early to the agent transition — you're already behind it.

Systems of record become worthless databases unless management moves first — incumbency is not a moat

$10 billion in cap table data doesn't protect Carta if tokenization bypasses it entirely. Murdock uses the company — one of his longest-held investments — as a live case study in how the same technology wave can make a business infinitely more valuable or render it a legacy database, depending entirely on execution speed.

"If tokenization of stocks come and they use Carta, then Carta is going to be infinitely more valuable... However, if tokenization comes and they bypass Carta... there's not much to say that that system of record is going to be worth much."

The same logic extends to Salesforce, which Murdock treats as an 8,000-meter peak — not melting overnight, but measurably less valuable if the ecosystem of companies built on top of it starts eroding. The tell isn't Salesforce itself; it's whether its application-layer partners survive the agent wave.

His framing is blunt: "What we have with the tsunami happening is a wake-up call to move to higher ground. Don't get caught on the beach when the damn thing hits the beach." Companies with rich contextual data can become dramatically more powerful if they deploy that context through agents. The ones that mistake data ownership for strategic safety will find out it isn't.

Per-seat SaaS is structurally broken — agents buy on consumption and the math doesn't work anymore

An autonomous agent doesn't care about your annual seat license. It will spin up 100,000 sandboxes in seconds, optimize probabilistically across vendors, and notify its human overseer when a spending limit is reached. That behavior demolishes the per-seat revenue model that underpins trillions in enterprise software valuations.

"If the agent has the access and has the authorization to use something, say a sandbox, you're just going to pay based on consumption. The agent will come and tell you, hey, we've reached our limit with the sandbox company here. What do we need to do?"

Murdock's portfolio company Docker has already been moving hard toward consumption-based pricing as it leans into AI. The E2B sandbox story makes the stakes concrete: human reaction time floors out around 400 milliseconds, which is why most mobile-era sandboxes were built to that spec. E2B responds in 80 milliseconds. No human notices the difference. Agents absolutely do — especially when they're running parallel workloads at massive scale.

The answer to which model, which tool, which vendor wins won't be made by a developer with opinions. "The answer to that question is going to be decided by the autonomous agent, not developers." Investors still modeling SaaS companies on seat-count growth assumptions are pricing a world that's already ending.

The open-source 'claw stack' will do for agents what LAMP did for the web — and create the platform opportunity of this cycle

In 2003 and 2004, post-9/11 austerity made Sun servers and Oracle databases unaffordable. The LAMP stack — Linux, Apache, MySQL, PHP — filled the vacuum and triggered the explosion of websites and e-commerce that Google rode to its 2004 IPO. Murdock sees autonomous agents hitting the same inflection point right now.

"I think in the same thing with autonomous agents that you're going to find the open source community coming out with the stack... they're going to have an orchestration layer where they can have multiple different LLMs that they can orchestrate on and triage workflows."

The orchestration layer is the key piece: agents routing workloads across Claude for quality-critical tasks, open-source models like DeepSeek or Llama 3 for cost-sensitive ones, dynamically optimizing at runtime. That routing function commoditizes the reasoning layer and shifts value upstream to whoever controls the orchestration standard — and downstream to the ASIC chip makers who will run those open-source models cheaply.

The closed model companies aren't the platform bet here. The open-source agent orchestration community is, and its growth rate is sheer numbers versus the concentrated resources of OpenAI or Anthropic.

Nvidia bought Groq to make CUDA viable for ASICs — Jensen knows the GPU era has a clock on it

The official story on the Groq acquisition was memory-on-chip performance. Murdock's read is more strategic and more threatening to Nvidia's incumbency. "He needs that capability so that he can handle ASICs chips eventually... I think the Groq acquisition is going to help them make sure that CUDA is viable for the ASICs explosion that's coming."

Meta's decision to say no to Jensen — betting its infrastructure buildout on ASICs instead of GPU clusters — is Murdock's exhibit A that the transition is real and being taken seriously by the largest buyers of compute on earth. When a company with Meta's capex appetite walks away from Nvidia's roadmap, Jensen notices.

The existential question for Nvidia isn't model commoditization. It's whether CUDA migrates cleanly to an ASIC-dominated world. If it does, Nvidia extends its moat. If it doesn't, the open-source agent stack runs on cheaper, more tunable chips and CUDA becomes a legacy layer. The Groq buy is Jensen's insurance policy against that scenario — an admission that the GPU era, while not over, has a visible horizon.

Tech private equity faces a Forstmann Little moment — and the firms that survive will be the ones who repriced their assumptions first

Teddy Forstmann went all-in on virtual telcos in 2000. Then 9/11 finished him off. Forstmann Little, once a defining buyout shop, was done. Murdock draws the line directly to tech PE firms today holding leveraged positions in legacy SaaS at 15-20x multiples.

"There'll be PE firms that end up like Forstmann Little... and then there'll be other ones that had the right bets and did the right creativity."

The metrics that used to anchor software valuations — revenue, growth rate, margin — have become, in his words, "transient slash questionable." A portfolio company's moat depends on whether its primary users could be replaced by agents within 18 months. Most PE models aren't stress-testing against that scenario. The firms that do the reckoning now, and reorient toward agent-native assets, survive and compound. The ones defending legacy portfolio theses won't get a second chance to adjust.

SMBs get disrupted first — and the 2028 election may be decided by autonomous agent job losses

Enterprise is always last. That's the pattern across every technology cycle, and Murdock expects the same here. The first wave of white-collar displacement won't show up in Fortune 500 workforce data — it'll show up in the hiring freezes and agent deployments at two-, three-, and four-person companies where a single autonomous agent replacing one executive assistant has enormous ROI.

"The first companies that are going to go with this are going to be small, medium businesses because they're the ones that, wow, one secretary makes a huge difference in their two, three, four person company."

The visible signal isn't mass layoffs — not yet. "The first thing you see is people stopping or slowing down hiring of white collar employees that input data and use a computer." Slower hiring doesn't make headlines. But it reshapes the labor market quietly, and Murdock is explicit that the political consequences will be loud: "That is going to be a major issue on the next presidential election. It could decide the election."

He floats minimum viable income — effectively an expanded UBI — as a likely policy response, with retraining programs layered on top. SMB-focused founders and investors have a narrow window before this dynamic is obvious to everyone.

Right now is the single best time ever to start a new fund — if you're willing to design it from scratch around agent buyers

Murdock's argument is tight: fund vintage timing explains more of venture returns than strategy. A 2005-2006 fund caught the mobile wave at the beginning; a 2009 fund missed Twitter, Facebook, and Uber entirely. The same logic applies now, except the sea change is larger.

"Absolutely the best time. You're in a sea change. Humans are no longer going to be the decision makers about software. It's going to be autonomous agents... people that embrace this new model can embrace it from scratch right now. You can have a huge advantage over people that have been successful with older models."

The structural edge for new managers isn't hustle — it's the absence of legacy assumptions. Established firms have portfolio companies built for human buyers, LP relationships anchored to metrics that are now "transient slash questionable," and partners whose pattern recognition was trained on a world that's being deprecated. A fund designed from day one around agent-native companies, consumption-based pricing, and agent-as-buyer dynamics carries no such baggage.

The real tell is who builds the agent orchestration standard — everything else follows from that

Cursor's obsolescence, Carta's fork-in-the-road moment, the Forstmann Little warning for PE, the SMB displacement wave — all of it flows from the same upstream event: the emergence of a stable, open-source autonomous agent stack. Whoever the LAMP equivalent turns out to be in this cycle, the companies that build on top of it early — and the funds that backed them before it was obvious — capture the generational return. The rest are optimizing legacy assets in a declining market. The tsunami doesn't ask whether you're ready.