The Situation
This Series A B2B SaaS company had a problem most growing startups would envy: more demand than they could build for. Their product-market fit was strong, their sales pipeline was full, but their engineering team — 8 engineers based in San Francisco — couldn't ship fast enough to capitalise on the moment.
Hiring locally wasn't an option. At SF engineering salaries, each additional hire would cost $220–280K all-in. The board had approved headcount growth, but not at that rate.
They'd tried offshoring twice before. The first attempt — a vendor in Eastern Europe — collapsed after 4 months due to communication overhead and inconsistent code quality. The second — a freelance team from Southeast Asia — lasted 6 weeks. The CTO told us: "I've written off offshoring as a category. But I'm out of options."
Why the Previous Attempts Failed
Before proposing anything, we did a post-mortem on both failed attempts. The pattern was identical: the offshore engineers were technically capable, but the communication model was broken.
Requirements arrived as Slack messages and Loom videos. Engineers asked for clarification; responses came 8 hours later. PRs were large, hard to review, and arrived without context. Feedback loops were slow enough that an engineer could spend a week building the wrong thing before anyone noticed.
This isn't an offshore problem. It's a communication latency problem. And Claude Code fixes it.
Building the Squad
We sourced 6 engineers from our network in Pakistan and Poland — markets where we've built strong pipelines of technically excellent engineers who are hungry to work with cutting-edge tooling. Every candidate went through a technical screen and a structured assessment of their written communication skills. Both matter equally for a Claude-Native squad.
Before the squad wrote a single line of production code, they went through our 4-week Claude Code certification:
- Week 1: Spec-driven development — reading and generating technical specifications, surfacing ambiguities before coding
- Week 2: Claude Code patterns for the client's specific tech stack (Next.js, PostgreSQL, tRPC)
- Week 3: Async communication discipline — how to write a PR description, how to escalate a blocker, how to give written feedback
- Week 4: Integration with the client's actual tools and a practice sprint on non-production features
The Onboarding
Week 5 was the real test: the squad's first production sprint alongside the SF team. We set a deliberately modest goal — 30 story points — and focused entirely on process, not output. We wanted to see how the squad communicated blockers, how they interpreted tickets, how they handled code review feedback.
They delivered 38 points. More importantly, the SF team's code review burden on squad PRs was lighter than on their own PRs — because every PR arrived with a Claude-generated summary, a test coverage report, and a list of edge cases considered.
The CTO's response after sprint 1: "These are the best-documented PRs I've seen in three years of running this team."
Sprint-by-Sprint Velocity
- Sprint 1: 38 points (target: 30)
- Sprint 2: 61 points
- Sprint 3: 84 points
- Sprint 6: 112 points — outperforming the SF team at 94
- Sprint 12: 118 points, consistent
The acceleration curve is typical of Claude-Native squads: slow start as engineers learn the codebase, then rapid compounding as their Claude Code patterns mature and their context of the system deepens.
The Results at 3 Months
Three months in, the squad had become the company's primary delivery team. The SF engineers shifted into architecture and product leadership roles — work that was both higher leverage and more satisfying.
Total engineering cost increased by only 35% while effective delivery capacity more than doubled. The squad costs 42% of what an equivalent SF hire would cost, with higher measured velocity.
The CTO, who had written off offshoring entirely, is now planning to grow the Claude-Native squad to 12 engineers by Q3.
Key Takeaways
- Previous offshore failures were communication failures, not talent failures — Claude Code fixes the underlying problem
- Written communication skills are as important as technical skills in hiring for a Claude-Native squad
- Certification before production code is non-negotiable — it sets the habits that everything else depends on
- The velocity curve compounds: Claude-Native squads get faster over time as their codebase context deepens
- The SF team got better too — freed from routine delivery, they did their best architecture work