Directions Brain

There are only two paths left for software

4 min read

There are only two paths left for software

Source: raw/articles/vc-articles/2026-03-23-a16z-two-paths-left-for-software.md

Note: captured from the public article URL using direct curl + pandoc fallback because the axcli chrome gateway was unavailable in this session. The raw file preserves the article body and provenance block.

Summary

david-george of a16z argues that public software companies have lost the “comfortable middle”: slowing growth plus incomplete true profitability will keep compressing equity value. He frames two credible paths for durable value creation: accelerate total revenue growth by 10+ percentage points through genuinely new AI-native products within 12-18 months, or rebuild the company toward 40-50% true operating margins including stock-based compensation. The article matters to this brain because its “living context layer,” token-path pricing, and output-over-platform arguments externally reinforce context-os-brain, agent-native-go-to-market, and vertical-use-case-led-brain.

Key claims

  • The middle is over for software. Public markets are no longer rewarding companies that are neither fast-growing enough for a premium multiple nor truly profitable enough for a fortress multiple.
  • Path one is AI-native growth. Incumbents need net-new AI products that can move total company growth by 10+ percentage points within 12 months; bolted-on chatbots or copilots attached to old SKUs do not count.
  • AI-native growth requires a company refounding. The article recommends process-capture sprints, harvesting SOPs/tickets/transcripts/requirements/policies/CRM notes/support logs/event data/approval paths, and turning documentation into a living context layer with evals for accuracy, exceptions, latency, and cost.
  • R&D should shift to small AI-native pods. The article recommends moving 50% of R&D to net-new AI products, using four-person pods, keeping PMs customer-facing, and keeping strongest engineers near core architecture to avoid fragmented product stacks.
  • Pricing shifts away from seats. The article argues that customers will look to take seat costs out as AI produces labor efficiency, while new growth sits in tokens, consumption, automations, outcomes, and machine-driven workflows.
  • Path two is true-margin restructuring. Companies without a credible AI-growth path should target 40-50% true operating margins including stock-based compensation through structural redesign, not cosmetic layoffs.
  • Engineering token budgets become operating leverage. The article treats high token spend per engineer as table stakes for productivity gains and argues organizations built around smaller, faster teams will outperform committee-heavy structures.
  • Old software moats are weakening. Data alone, integrations, workflow, and UI advantages become less defensible as agents get better at moving across systems and migration becomes easier.
  • Broadcom/VMware is the margin-path example. The article uses Broadcom’s VMware simplification and adjusted EBITDA guidance as evidence that radical cost discipline and product simplification can work, while warning it is not a cultural blueprint for every founder.
  • The board-level question is binary. Founders and investors should ask whether the company is on the AI-growth path or the true-margin path; “a little of both” should keep attracting market pressure.

Hypotheses raised

Notable quotes

  • “the comfortable middle is over.”
  • “accelerate revenue growth by 10+ percentage points”
  • “rebuild the company to 40%+ true operating margins”
  • “Create a living context layer, not a pile of static PDFs.”
  • “If you are not in the token path”
  • “No middle lane.”

Mentioned in

Graph View

Backlinks