Uber's 95% AI Adoption Is a Sourcing List. June 3 Removed the Guards.
Uber cut 23% of People and Places days after capping Claude Code at $1,500. Here's how to source the leaderboard cohort before Anysphere does.
On June 3, 2026, Uber cut 23% of its People and Places division, the team that owns recruiting, retention, and counter-offers. Three weeks earlier, CTO Praveen Neppalli Naga had publicly admitted Uber burned its entire 2026 agentic coding budget in four months, then capped Claude Code and Cursor at $1,500 per engineer per month. If you hire AI-native engineers, those two events combine into the cleanest 30 to 90 day window of the year.
What actually happened, in order
In December 2025, Uber rolled out internal access to Claude Code and Cursor and started ranking engineering teams on an internal leaderboard by AI tool usage. Adoption went from 32% to 84% between rollout and February. By the April disclosure, 95% of Uber's roughly 5,000 engineers were using AI tools monthly, around 70% of committed code was AI-generated, and an internal agent was landing roughly 1,800 code changes per week with no direct human input.
Then the bill came in. Neppalli told The Information in April: "I'm back to the drawing board because the budget I thought I would need is blown away already." He himself had consumed $1,200 in tokens during a two-hour internal demo. Heavy users were billing $500 to $2,000 a month. The $1,500 per-tool soft cap followed, with tiered thresholds and an exception dashboard. An engineer running both Claude Code and Cursor can spend up to $3,000 before throttling, but the heaviest agentic workflows already exceeded that.
On May 26, COO Andrew Macdonald went on the Rapid Response podcast and said the quiet part out loud about the ROI: "That link is not there yet" between rising Claude Code spend and shipped consumer features.
Eight days later, Jill Hazelbaker, newly promoted to president and chief corporate affairs officer per Uber's 8-K, ran the cuts. Her background is comms and policy (Snap, Google, the Bloomberg campaign, the McCain campaign), not HR. Bloomberg and CNBC both flagged that many of the eliminated People and Places roles were senior. Uber insists the cuts are unrelated to AI and represent "well under 1%" of its 34,000 corporate employees.
That denial is the tell. The cuts removed senior recruiters, the exact people who run counter-offers and exec save plays, in the same quarter the COO publicly questioned whether the AI investment is shipping anything.
The 95% number is a filter, not a flex
Most sourcers will read "95% of Uber engineers use AI tools monthly" and treat it as a marketing line. It isn't. It's a population definition. Uber publicly tracked Claude Code and Cursor usage on internal leaderboards. The top quartile of that leaderboard is identifiable from the outside if you know where to look:
- Talks and posts on eng.uber.com referencing agentic workflows, MCP servers, or in-house agent infrastructure.
- GitHub activity on agentic SDKs, MCP server repos, Cursor extensions, and Claude Code wrappers.
- Conference talks at AI Engineer Summit, KubeCon, and QCon naming internal tooling.
- Cursor community Discord and forum presence, plus Anthropic developer forum activity.
- Co-authorship on internal RFCs that leak into recruiter-facing portfolios.
This is exactly the kind of search that breaks LinkedIn Recruiter. The signal lives across GitHub commits, conference speaker pages, blog bylines, and forum handles, not in the "Skills" field. Which is why we built Refolk: you describe the engineer in plain English ("Uber backend engineer in the Bay Area with public Claude Code or MCP server contributions in the last 12 months") and get a ranked shortlist that actually reflects the cross-source evidence.
Why the recruiter cuts matter more than the budget cap
The budget story will get the headlines. The recruiter cuts will move response rates.
Senior in-house recruiters do three things that matter to sourcing math: they detect early flight risk, they run aggressive counter-offers, and they coordinate exec save plays for staff and principal ICs. Bloomberg's reporting and Uber's own SEC filing make clear that many of the eliminated roles were senior. Hazelbaker is a comms executive, not a people executive. The org running retention right now is structurally different from the one that existed in May.
In practice, that means three things for outreach in June and July:
- Reply rates on cold outreach to Uber staff and principal engineers will rise. The internal retention machine is understaffed and disorganized.
- Counter-offers will be slower and less aggressive. First offers from outside will close faster than they did in Q1.
- The window is finite. Meta, Anysphere, and the Codex-adjacent labs will work the same list. The clean pulls happen in the first 60 days.
Uber loudly denies AI caused the cuts. It also just removed the people whose job is to keep its AI-fluent engineers from answering your email.
The pitch is judgment, not throughput
Generic "we use AI" outreach will get ignored by this cohort. They live inside a leaderboard culture and they heard their own COO say the link between token spend and shipped features "is not there yet." They know.
The pitch that converts here is judgment over throughput. Lead with shipping to production, not to a leaderboard. Name the tradeoff explicitly: at Uber they were ranked on tool usage; at your company they'll be measured on customer outcomes. If you're a startup offering unlimited Claude Code, Codex, or Cursor spend, say that in the first line. An Uber engineer who was running $2,000-a-month parallel agent workflows in March just got throttled to $1,500 per tool. The comp delta is now legible to them in dollars.
The Microsoft and Walmart context strengthens this. Microsoft is forcing its Experiences and Devices division (Windows, Microsoft 365, Outlook, Teams, Surface, all under Rajesh Jha) off Claude Code and onto GitHub Copilot CLI by June 30, 2026. Walmart capped its internal assistant. Across the enterprise, the engineers who kept their full Claude Code plus Cursor plus Codex stack are a scarcer profile than the raw headcount suggests. Uber is one of the few remaining shops where engineers have current, simultaneous, production-grade exposure to all three.
Where the addressable pool actually lives
Refolk's index returns roughly 1,642 current U.S.-based Uber software engineers, staff engineers, and engineering managers. The geographic concentration is sharp:
- San Francisco Bay Area (SF, San Jose, Sunnyvale): the dominant cluster, where the agentic tooling teams sit.
- New York: secondary concentration, heavier on product and growth engineering.
- Seattle: smaller but meaningful, with infra and data engineering depth.
That is a finite list. Against it, you have a known competitor set. Anysphere (Cursor) has been pulling senior PM and engineering talent from GitHub and Microsoft throughout 2026, and Uber engineers are a natural landing pad given their daily Cursor exposure. OpenAI is hiring Codex-adjacent talent as Uber adds Codex to its agentic stack. Meta is rebuilding after its own May cuts.
For sourcing Uber engineers in this window, the unit economics are unusually friendly. The pool is bounded, the retention machine is gutted, the comp pitch writes itself, and the qualifying signal (verifiable agentic-tool depth) is sitting in public.
A 30-day playbook
If you're running point on sourcing Uber engineers right now, here's the sequence that works:
Week 1: build the leaderboard cohort
Pull the named engineers from eng.uber.com posts in the last 18 months, cross-referenced with GitHub activity on agentic infra. Layer in conference speakers from AI Engineer Summit and QCon who list Uber as their employer. This is the search where Refolk's plain-English query outperforms boolean strings on LinkedIn, because the qualifying evidence is spread across four sources. Tech layoffs in 2026 are running at roughly 974 per day, 44% above 2025's pace per TrueUp, so general sourcing supply is up, but this specific cohort is differentiated by verifiable depth, not availability.
Week 2: lead with the token-budget pitch
For startups with unlimited agentic spend, lead the first message with the dollar figure. "We pay your tokens, no cap" is a concrete benefit an Uber engineer can price against the $1,500-per-tool soft cap. For larger employers, lead with judgment-over-throughput framing and name the Macdonald quote.
Week 3: target staff and principal first
Senior recruiters at Uber were the ones cut hardest. Counter-offer machinery for staff and principal ICs is slower than it was 90 days ago. Hit that band first while the window is open. AI-fluent engineer sourcing at the staff level converts at a different rate than at L4 because the comp deltas are larger and the retention conversations are more political.
Week 4: close before the People and Places team rebuilds
Hazelbaker will rebuild eventually. The clean window closes when she hires a head of talent. Move offers in 14 to 21 days, not 45. If you're trying to poach Uber engineers in 2026, treat June and July as the only quarter where the math is this favorable.
number: $3,000
label: Maximum monthly spend an Uber engineer can hit before throttling, across Claude Code and Cursor combined
note: Heavy users were already billing $500 to $2,000 per tool. Unlimited-spend offers from startups are now legible in dollars.