Approxima
Your software should build itself.
Winter 2026ActiveB2BEngineering, Product and DesignArtificial IntelligenceEnterprise SoftwareInfrastructureSan Francisco, CA, USA
Company
https://approxima.aiHuman-assisted development should be for novel ideation, not routine maintenance, bug fixes or incremental updates. Approxima is bringing you that future with agents that can scope work, code, validate their fixes end-to-end in sandbox environments and merge to production. Book a demo: https://calendly.com/ashish-selvaraj/approxima
Verdict
High Signal
Market Opportunity
Autonomous software development / AI coding agents is a massive B2B market — enterprises spend hundreds of billions on software maintenance annually. The ICP (engineering teams wanting to automate routine bug fixes, maintenance, and incremental updates) is clear and real. Monetization path is obvious via seat-based or outcome-based pricing to dev teams.
Medium Signal
Founder Signal
Both are UWaterloo CS '25 grads — fresh graduates with strong internship pedigrees. Ashish has co-op stints at Databricks (pipelines execution engine in Scala), Microsoft, and theScore, plus a self-built app (Not Evil Sudoku, 180K installs) showing indie product chops. Ethan interned at Citadel Securities, Jane Street, and Squarepoint — elite quant finance shops, impressive signal but less directly relevant to autonomous coding agents. Neither has shipped a production startup at scale or had a previous exit; this is their first company.
Low Signal
Competition
This space is brutally crowded: Devin (Cognition AI), GitHub Copilot Workspace, Cursor, Codegen, SWE-agent, OpenHands, and numerous others are all attacking autonomous code agents. Google and Microsoft (via GitHub) have massive distribution advantages. No proprietary data moat or technical differentiation is visible from public materials — the description reads generically similar to every competitor's pitch.
Low Signal
Product
The website is literally just a login page with no public-facing product, marketing copy, feature descriptions, or social proof. The only call to action visible is 'Send Magic Link' — zero product substance is visible. Description mentions agents that scope, code, validate, and merge to production, but nothing is demonstrable or publicly accessible.
OverallC Tier
Two technically credible UWaterloo fresh grads with strong internship backgrounds, but zero visible product traction, no press, no customer logos, and a completely opaque website. They're entering one of the most competitive spaces in AI — autonomous coding — where well-funded incumbents like Cognition (Devin), GitHub Copilot Workspace, and Cursor already have significant distribution. The 180K installs on Ashish's side project is the only real traction signal in the entire application. Without evidence of paying customers, a differentiated technical approach, or any product surface to evaluate, this reads as two smart new grads with a compelling idea and no demonstrated moat.
