Software development market insights

Coding agents went mainstream this summer, and the leaderboard flipped

Most professional developers now hand entire tasks to AI agents, and the tools that led a year ago are no longer in front. Two independent studies published in August confirm it: JetBrains Research, which tracks tool adoption across its global developer base, and GitKraken's State of AI in Engineering 2026, a survey of developers and engineering leaders.

• 90% of professional developers use AI coding agents at least weekly. 68% use them daily.
• Claude Code reached 39% adoption globally and 47% in the US, roughly double GitHub Copilot, which fell from 29% to 21% in a year.
• 28% of developers now delegate work to agents as their primary workflow, up from 7.6% in September 2025.
• 47% of enterprise developers run agents all day, compared to 32% at smaller companies.

Everyone feels faster, but it's still hard to prove

Developers report large productivity gains from AI, but most companies have no data to back it up. GitKraken's 2026 survey calls this the proof gap, and the numbers are blunt.

• 84% of developers say AI makes them more productive. 43% say "much more."
• Only 20% of companies measure that impact with specific metrics. 39% have no measurement capability at all.
• 72% of organizations rely on belief or self-reporting rather than data when judging AI's effect on engineering output.
• 96.4% of teams already use AI coding tools. Adoption is no longer the differentiator. Measurement is.

Engineering talent is concentrating in fewer companies

Job postings are up and AI engineering demand is climbing, but the talent is concentrated in a few companies. That's why more CTOs are adding partner capacity instead of waiting on a job posting. The Pragmatic Engineer's State of the Job Market 2026, Gergely Orosz's analysis of industry headcount and job posting data, shows where the growth is going.

• Software engineering postings rose in the US and UK. Top tech companies are hiring roughly 20% more developers than a year ago.
• AI engineering listings at Apple, Google and TikTok grew 50 to 100% year over year.
• Fastest-growing engineering orgs are in fintech, security and observability: Ramp +94%, Wiz +84%, Datadog +68%.

Highlights from the past quarter

OAK'S LAB is expanding to San Francisco

Theo and I grew up in the Bay Area and started OAK'S LAB in Prague ten years ago. We're moving back to San Francisco to establish a physical OAK'S LAB presence and meet the CTOs and founders we've been working with from thousands of miles away. If you'd like to meet, or know someone in your network we should be talking to, reach out. We'd be happy to talk.

Three products launched in Q3, all on time, all without a single production issue

Three separate teams shipped three major launches this quarter, each on the date committed and without a critical issue in production. It's the standard our clients expect from us, and the reason CTOs at some of the fastest-growing companies keep trusting us with their most important releases.

• A consumer iOS app with an AI investment agent for a private markets platform that has facilitated $80B+ in liquidity for 200,000+ participants
• An identity and trust verification app for hiring, the "TSA PreCheck for employment," shipped to iOS and Android for a cybersecurity startup
• An adaptive AI exam-prep platform for an education brand trusted by 100,000+ families: diagnostic testing, personalized daily practice, mock exams, and an AI study partner that tutors through Socratic questioning

Our hiring bar just got higher: 0.3% of candidates make it in

Our people are our product, and the teams we compete with hire from Google, Microsoft and the fastest-growing startups. To win that comparison, we have to be more selective than they are. Year to date, our acceptance rate is 0.3%, down from 0.4% in the first half of the year. Every candidate goes through multiple rounds covering technical skills, live problem-solving, and culture fit, so the people we place into your product org are difference makers from day one.

Featured portfolio client

PlexTrac

In August, PlexTrac was acquired by Brinqa, and founder Dan DeCloss joined Brinqa's executive team and board. We've been part of that story since the beginning. In the early days, when Dan came to kick off the project in Prague, PlexTrac was a bootstrapped MVP with one internal engineer. Our team rebuilt the architecture and UI for enterprise buyers, shipped the product modules that landed customers like Mandiant, and stayed embedded through a $10M Series A and a $70M Series B before handing off to Dan's own team in Boise. Watching a founder we believed in from the first sprint reach an exit is the reason we do this work. Congratulations, Dan.

Read the case study

Portfolio clients in the news

Blackpoint Cyber

Launched new identity threat detection and response (ITDR) capabilities in its CompassOne platform - Blackpoint Cyber (September)

Blue Owl Capital

Unveiled plans to take a data-center REIT public, seeded with ~$6.5B of its own assets, to fund AI-infrastructure buildout - Yahoo Finance (September)

OpenLoop Health

Acquired Hey Revia, an AI-powered patient communication platform, and appointed its first Chief Technology and Product Officer to scale its virtual care infrastructure - MobiHealthNews (July)

Next available team opening: November 1st

An exceptional seven-person product development team becomes available at OAK'S LAB next month.

This is a cross-functional squad that has worked together for years, shipping software for US growth-stage startups from Series A to D. They report directly to the CTO, and one client called them a "delta force" that gets deployed against the company's hardest product problems. They can own an entire product or vertical end to end, from discovery through delivery, using an AI-forward development process that puts agents to work without giving up control of the codebase.

No storming, no norming. High-performing from day one.

Team composition

• Product Manager, 10 years of experience
• UI/UX Design Lead, 11 years
• Tech Lead, 13 years
• AI Architect, 11 years
• 2 Full-Stack Engineers, 11 years average
• QA Engineer, 12 years

Core stack: React, Next.js, Node.js, TypeScript

Why OAK'S LAB

• We partner with only 10 clients at a time, every client gets attention of senior leadership
• We focus exclusively on US growth-stage tech companies (Series A to D)
• Deep domain expertise in cybersecurity, fintech, healthcare and education
• 80 products shipped over ten years, founded 2016
• We hire the top 0.3% of engineering talent
• ISO 27001 certified

Last quarter through snapshots