Overview of 20VC with Jerry Murdock
This episode features Jerry Murdock (co‑founder of Insight Partners) in conversation with Harry Stebbings about the seismic industry shift driven by autonomous AI agents. Jerry argues that autonomous agents—not just "AI" generally—are the coming tsunami that will reorder software, chip economics, labor, venture/private equity, and product strategy. He draws on three decades of investing to explain where value will move, which companies are well‑positioned (or at risk), and what founders, investors and executives should do now.
Key themes & high‑level takeaways
- The big disruption is autonomous agents (agents that can act, triage, and write code), not just better LLMs. These agents will change how software is built, bought and run.
- Open‑source agent communities + orchestration layers will be central: expect a new "agent stack" (analogous to LAMP) that coordinates multiple LLMs, sandboxes, reasoning, and runtime.
- Chips will migrate from general GPUs to specialized ASICs (and memory‑on‑chip designs) for cost/latency advantages. Who executes on this migration matters.
- Systems of record can either become far more valuable (if they integrate with agents/tokenization) or marginalised (if bypassed). Execution decides.
- Pricing and go‑to‑market models will shift to consumption (agents as employees with credentials => consumption billing; agents negotiating purchases).
- Labor displacement among white‑collar roles (exec assistants, customer support, junior devs, bookkeeping, parts of legal/healthcare) will be material and politically significant; universal basic income / minimum viable income may reappear in policy debates.
- Timing and bold bets still matter in VC; the current period is prime for new funds and for AI‑native startups.
Topics discussed (with summary)
Autonomous agents as the "tsunami"
- Agents are probabilistic decision‑makers that can orchestrate multiple models, spin up many sandboxes, and even write production code.
- Early AI‑native startups are using agent tooling to generate and maintain code — that quickly changes the value proposition of coding platforms and developer tools.
- Many founders/investors see incumbent developer tools (e.g., Cursor in the transcript) as potentially obsolete unless they embrace agents rapidly.
Open‑source community and the "agent stack"
- A community‑driven stack (reasoning, orchestration, sandboxing, model routing) will emerge. Orchestration decides which model to run for each workflow.
- Orchestration + routing -> ability to blend expensive proprietary LLMs and cheaper open‑source models for cost/quality tradeoffs.
- This dynamic accelerates adoption of optimized hardware (ASICs) and specialized chips with memory‑on‑chip.
Chips and infrastructure
- Expect migration pressure from NVIDIA GPUs (CUDA) to ASICs optimized for specific model workloads. Memory on chip and cost per inference will be key.
- Strategic acquisitions (e.g., “Grok” referenced in the interview) are viewed as moves to prepare for chip heterogeneity and on‑chip memory needs.
- Whether NVIDIA retains value depends on execution (migration strategy, CUDA portability, partnerships).
Systems of record & platform incumbents
- Systems of record (e.g., Carta, Salesforce) have two possible futures: leverage their context/data to become agent platforms and thus increase in value, or be bypassed by new systems designed for agents and decline.
- Incumbents with ecosystems (third‑party apps built on top) are more resilient but must adapt quickly.
Business models and pricing
- Agents will act like employees: they have identities, credentials and authority. Billing will be consumption‑based (compute/sandbox/memory usage).
- Product teams must prevent "product debt" by validating upstream (use Reforge‑style discovery) before shipping; shipping faster isn’t useful if agents/end users don’t use it.
Labor, society and politics
- Initial displacement will hit roles that primarily input/manipulate data: customer support, bookkeeping, legal assistants, junior developers.
- SMBs likely to adopt agents earlier (high ROI from automating a single human role); enterprise adoption follows but more slowly.
- The political response (e.g., guaranteed minimum income, retraining programs) could become a central election issue.
Venture, private equity and fund strategy
- Timing is still a dominant determinant of fund performance. New funds that embrace the agent era can gain outsized advantage.
- Some private equity firms will struggle if they don’t reassess assumptions; others will thrive by pivoting and deploying agent‑powered diligence and analytics.
- Jerry’s investing posture: go big on differentiated, high‑momentum ideas (his Twitter investment as example).
Notable quotes
- "The tsunami is about autonomous agents, not just AI in general."
- "Move to higher ground. Don't get caught on the beach when the damn thing hits the beach."
- "Agents are employees: you give them credentials, you give them identity."
- "Money does not come with instructions."
- "You don't know the edge unless you go over it."
(Transcript includes some garbled names/terms — e.g., "OpenClaw"/"NanoClaw" and "ChatTPT" — the intended meaning appears to be open‑source agent/LLM ecosystems and ChatGPT/LLM families.)
Actionable recommendations
For founders / product teams
- Build for agents: design APIs, credentials, and usage billing that an agent can call autonomously.
- Validate upstream (customer discovery) before shipping features — agents amplify wasted engineering if product/market fit is missing.
- Instrument data and context now — systems of record that have usable context for agents will be defensible assets.
For VCs / investors
- Invest in orchestration, sandboxing, agent tooling, efficient inference hardware (ASICs), and AI‑native apps that are agent‑first.
- Use autonomous agents internally for sourcing, diligence, and market analysis to level‑set competition.
- Expect higher dispersion: most investments will underperform; a small set of hits will drive returns.
For incumbents / enterprise leaders
- Experiment now with autonomous agents in low‑risk workflows (sandboxes, internal automation) but prepare for production adoption.
- Reevaluate revenue models (consumption billing, pay‑per‑action) and contract terms to accommodate agent purchasing.
For policy / public leaders
- Anticipate labor displacement among white‑collar roles and create retraining, safety nets or minimum income conversations now.
Quick checklist (90‑second)
- Are you agent‑ready? (APIs, credentialing, sandboxing)
- Do you have instrumented data and context for agents to act on?
- Is your pricing flexible for consumption billing?
- Are you tracking orchestration or model‑routing as a strategic capability?
- For investors: have you doubled down on open‑source/agent ecosystems and hardware plays?
Final notes
- Jerry emphasizes execution and adaptation: the structural change is happening fast but outcomes depend on who executes best.
- He believes this is a once‑in‑a‑generation time to found funds and startups oriented around agents.
- Moral/personal themes: iterate fast, accept failures as learning, and respect money as a form of energy to be used intentionally.
If you want, I can extract a one‑page investor memo from this episode (thesis + target sectors + representative companies to track) or a founder checklist tailored for building agent‑first products.