AI Coding Tools — Market

Verified claims and product-axis read for AI Coding Tools. Every fact below is sourced; every product judgment traces back to underlying signals.

Verified facts

• Anthropic's Claude Sonnet 4.5 and 4.7 model cards specifically benchmark 'long-horizon coding' separately from chat ↗ (other)
• Claude Code added background tasks / scheduled agents in late 2025, blurring CI and IDE workflows ↗ (other)
• Replit removed its 'always-on repls' free tier in 2024 to shift compute spend toward Agent users ↗ _(historical_event)_
• Cursor publicly stated it serves >1B model requests per day by mid-2025 ↗ (other)
• Augment Code published whitepaper claiming 40%+ developer-time savings on Fortune 500 pilot ↗ (other)
• Cursor's free tier ARPU conversion rate to Pro is reportedly >5%, well above SaaS norms ↗ (other)
• Anthropic's revenue mix shifted to ~80% API / 20% consumer in 2025, with coding the largest API workload ↗ (financial)
• Cursor's seed deck (2022) framed the company as 'AI-first IDE', not 'AI assistant in IDE' ↗ (other)
• Cursor's enterprise pricing went from public list to 'contact us' in 2025 ↗ _(historical_event)_
• Open-source coding model leaders shifted from Code Llama to Qwen2.5-Coder and DeepSeek-Coder by 2025 ↗ (other)

Top products (engine read)

Replit Agent — browser-hosted full-stack app-builder agent

Opportunity: Non-developer / beginner audience that wants a prompt → running app loop without leaving the browser; backed by $100M ARR run-rate in ~12 months.

Hot — Replit is hiring an Agent Platform staff line, an entire Cloud/Compute infra org, and a full enterprise GTM (AE, SE, Sales Director, Investment Associate), classic late-stage scale-up shape. Bet: capture the prosumer 'vibe-coding' wedge before Cursor/v0 cannibalize it.

Cursor — AI-first IDE

Opportunity: Default IDE-shaped AI workflow that swaps in best-of-breed frontier models within weeks and an agent loop (Composer) that now drives >40% of sessions, pulling Pro user token consumption 5x YoY.

Reigning prosumer AI IDE: $9-10B valuation, SOC2+SAML unlocking F500, but visibly leaking power-users to Claude Code CLI per HN chatter; defensibility is product-velocity, not moat.

Claude Code — agentic coding CLI

Opportunity: Frontier-model CLI that bypasses IDE chrome; backed by 1M-context tier for large-repo agentic loops and >50% of Anthropic API revenue coming from coding workloads. Capacity is the binding constraint.

Currently the wind direction for power-users on HN — moving the unit-of-work from autocomplete to a multi-hour agent. Risk: throttling and unit-economics of flat-rate Pro/Max plans threaten user trust.

GitHub Copilot — IDE-embedded AI assistant

Opportunity: Distribution incumbent fighting on two fronts: (1) defending VS Code/repo footprint while Claude Code & Cursor poach power-users, (2) shifting from flat-fee to token-metered Enterprise (per 8d3d84dd) without alienating the base.

Distribution-dominant but losing narrative — HN sentiment ranges from skeptical to hostile, and pricing churn (Pro+ tiering, metered code-review) suggests margin pressure. Still the default in enterprise procurement.

See the Products and Strategy modules for the full product list and forward-looking judgment.