VP Engineering at an F500 tracking real developer productivity gains, tool sprawl, mandate enforcement.

Audience Profile

• Age / Experience: 12-25 years
• Current role: VP Engineering / SVP Engineering / CTO of platform
• Top pain points:
• Real productivity gains from AI tools are smaller than vendor claims
• Tool sprawl as each team picks different AI assistants
• Mandate compliance theatre (engineers route around official tools)
• Top decision blockers:
• No standardised benchmark for AI-assisted engineering productivity
• Mandate-vs-judgment tension on top-down tool selection
• Engineer-side cynicism about vanity KPIs erodes trust in the mandate

What This Segment Needs

• Information: Honest productivity benchmarks (cycle time, defect rate, not LOC)
• Tools: Cursor vs Cognition vs Copilot enterprise comparison
• Services: Mandate-design frameworks that avoid Goodhart traps

Top 5 Companies for You (Fit Score)

| Rank | Company | Score | Why | |------|---------|-------|-----| | 1 | Databricks | 84/100 | Run rate $3.7B (2025-06-11) → $4B+ (2025-09-16), ~50% YoY; $1B AI-product ARR; Series K $1B+ at ~$100B val. Mosaic AI research-to-product pipeline. Counter: ~25x ARR, sales-heavy 2026 GTM postings. | | 2 | Cursor | 83/100 | ARR ~$500M → ~$1B (2025-11-13), Series D ~$29.3B val; engineers at >half Fortune 500. Composer in-house model, Background Agent GA. Counter: ~29x ARR, leans on OpenAI/Anthropic compute who ship rival agents. | | 3 | Snowflake | 83/100 | Product revenue $996.8M → ~$1.2B (Q1→Q3 FY2026, ~+30% YoY); >$1M customers 606→654; weekly AI accounts 5,000→6,100+. Public SEC filer, $6.7B RPO. Counter: GAAP-unprofitable, NRR 124–125%. | | 4 | Cognition | 82/100 | $400M raised 2025-09-12 at $10.2B post; Windsurf acquired ($82M ARR, 350+ customers); Goldman piloting Devin vs ~12,000 devs; SWE-1.5/SWE-grep models. Counter: Google poached Windsurf CEO days prior; only ~$82M ARR. | | 5 | OpenAI | 81/100 | WAU ~700M → 800M (2025-10-06 DevDay), 4M developers; AgentKit + GPT-5. Counter: ~$400B near-term compute commitments, private/unprofitable, financing-dependent; CTO of Applications reshuffle 2025-09-02. |

Deal-Breakers (Your Hard Preferences)

No hard preferences declared for this segment.

How to Evaluate Any Company in this Niche (Checklist)

• [ ] Check growth signals: require two dated run-rate/ARR disclosures 6+ months apart with a YoY %, not one vendor figure.
• [ ] Check comp data: none of these 5 disclosed comp — pull levels.fyi band for the target title; treat as unknown until offer stage.
• [ ] Check learning signals: confirm a funded applied-research req (e.g. "Research Scientist, Foundation Models") plus a Staff/Principal IC ladder, not GTM-only postings.
• [ ] Check stability signals: compute valuation ÷ ARR; flag >20x with undisclosed profitability as a burn/down-round risk.
• [ ] Check culture signals: ask in interview "what % of eng uses the mandated AI tool weekly, and measured how?" — the Goodhart test.
• [ ] Check mandate fit: ask whether productivity is tracked by cycle time / defect rate or by LOC / acceptance-rate vanity metrics.

Reverse-Hype Watch

• Cursor's Composer "first in-house frontier coding model" / sub-30s agent-turn claim is vendor-stated with no independent benchmark.
• Cognition's SWE-1.5 (~13x speed) and SWE-grep (~20x faster context) are self-reported; the flagship Goldman Sachs Devin deployment is still a scaling pilot, not booked revenue.
• OpenAI's compute commitments (~$400B near-term, Stargate ~10GW, AMD 6GW) outrun disclosed economics while the company is private and unprofitable.

What's under-reported for this segment: every signal above proves these vendors are *bought*, never that they make engineers *faster*. The dimension you actually own — measured cycle-time delta, defect-rate change, and the route-around rate of mandated tools at F500 scale — is essentially absent from public coverage, because vendors publish ARR and adoption logos, not independent productivity methodology. Assume any "X% gain" without a named control group and metric is marketing until proven internally.