On October 7, 2026, Fortune reported that Dropbox CTO Ali Dasdan took AI tool adoption across engineering, product, and design from under 40% to full coverage, and that non-technical staff now ship their own Lovable apps. If your resume still leads with a FAANG line and a Skills row that reads "Claude, Cursor, Copilot," you are optimizing for the screen that just stopped mattering. The new screen looks for artifacts, and most candidates do not have any on the page.
The screen changed from pedigree to builder's disposition
Hiring managers at AI-native teams are now scanning resumes for evidence that you make things on your own, because years of experience showed zero correlation with AI adaptability in work trials. The widely shared "new currency" essay from Tiang Lim put it plainly: cool personal websites, side projects, and an obvious love of making predicted success, while FAANG pedigree and name-brand universities predicted almost nothing.
"Builder's disposition" is a disposition screen, not a skills screen. That is the part candidates are getting wrong. People read the trend, panic, and cram four model names into their Skills section. The screen is looking for a URL, a repo, or a shipped Lovable app, not a keyword. If the top of your resume reads like a LinkedIn endorsement cloud, you are failing a test you did not know was being administered.
The mechanism is simple. When roughly 70% of code at Dropbox and Uber is AI-generated, "shipped X features" bullets are deflated across the entire candidate pool. The scarce evidence is judgment: evals you built, hallucinations you caught, a cheaper small model you swapped in for a Claude call. Anyone can list the tools. Few can point at the thing they built with them last weekend.
What Dasdan actually did, and why the "mandate" frame is wrong
Dasdan did not mandate AI adoption top-down at Dropbox's 2,113-person company. He made tools widely available, ran AI hackathons, championed internal superusers, and published metrics on how Claude, Codex, and Cursor sped up work until adoption hit 100%. Companies now hiring from this playbook do not want candidates who complied with a mandate. They want the bottom-up adopters who were already doing it before one existed.
That distinction matters for how you phrase bullets. "Adopted Cursor per company rollout" is a compliance bullet. "Ran a Friday AI hackathon that produced two internal tools, one of which replaced a $14k/mo vendor" is a disposition bullet. One sounds like a passenger. The other sounds like the person Dasdan would have promoted to AI superuser.
Dropbox's own tech blog describes Nova, an internal service for running coding agents, and Cursor indexing a monorepo of more than 550,000 files against infrastructure serving over 300,000 requests per second. These are not features Dropbox bought. They are things Dropbox people built because the tools were lying around and the disposition was there. Your resume should name the equivalent in your own work.
From work-trial tasks intentionally too long to finish by hand, per Tiang Lim's "new currency" essay.
The Cursor vs. Copilot signal, by the numbers
In Refolk's index of U.S. professional profiles, 9,114 people list GitHub Copilot as a skill versus 1,360 for Cursor, a 6.7x gap, but Copilot's top employers are legacy enterprises like Optum, Spectrum, and Capstone Logistics, while Cursor's top titles are Founders, Co-Founders, and Staff AI Engineers at stealth startups. If you are applying to an AI-native team and your resume lists only Copilot, you are flagging yourself as late to the curve.
| Signal on resume | U.S. profile count | Dominant titles | What it signals |
|---|---|---|---|
| Lists "Cursor" | 1,360 | Founder, Co-Founder, Staff AI Engineer | Scarcity, frontier-team adjacency |
| Lists "GitHub Copilot" | 9,114 | CTO, Principal/Staff SWE at legacy enterprise | Mass-market, lagging |
| Copilot : Cursor ratio | 6.7x | n/a | Copilot is now baseline, not differentiator |
| Dropbox eng/product/design AI adoption | <40% to 100% | n/a | The pattern now spreading |
| AI-fluent vs. AI-aware trial gap | ~10x | n/a | Why "familiar with" bullets fail trials |
The mechanism behind the Copilot lag is adoption cascade. The tool a company's champions used two years ago is the tool today's resumes list, because resumes trail actual usage by about one promotion cycle. Frontier teams moved to Cursor and Claude Code while the median resume caught up to Copilot. If you are writing for 2026, your tools section needs to show you are already one step past the median.
Per a Nexford University study cited in Built In, 26% of hiring managers now treat AI fluency as a baseline job requirement and 35% specifically look for AI experience or coursework on resumes. "Baseline" is the operative word. Baseline means you do not get credit for having it. You lose points for not having it. The upside bullets are the ones above baseline: shipped artifacts, evals, cost tradeoffs.
Rewrite your bullets around judgment, not usage
Replace tool-name bullets with judgment bullets that show you picked between models, caught a failure mode, or swapped a cheap small model in for a frontier one. Fortune reported in August 2026 that leaders are now capping AI spend and retraining staff that smaller cheaper models can often do the job, which means cost-aware bullets have moved from clever to required.
Here is the before/after pattern that reads correctly in late 2026:
- Before: "Used Claude and Cursor to accelerate development."
- After: "Shipped a Cursor-driven refactor of 42k LOC in one sprint; built eval suite that caught a hallucinated auth check before merge."
- Before: "Familiar with GPT-4, Claude, Copilot."
- After: "Replaced a Claude Sonnet call in our support pipeline with a fine-tuned 8B model; cut inference cost 71% at matched accuracy."
- Before: "Led AI adoption on team."
- After: "Ran weekly office hours that moved team Cursor usage from 2 of 11 to 11 of 11 in six weeks; published internal metrics dashboard."
Each after-bullet contains a number, a tool, a decision, and a result. That is the structure the new screen rewards because it demonstrates the thing years-of-experience failed to predict in trials. Tailoring every posting's bullets by hand is the work Refolk takes off you: paste the posting, get your resume back rewritten against it with the judgment bullets surfaced and the keyword clutter cut.
The screen is looking for a URL, a repo, or a shipped Lovable app. Not a keyword cloud.
Add a "Shipped" section above your work history
Create a new section called "Shipped" or "Built" above your experience, listing three to five live links to things you made: a Lovable app, a Cursor-built internal tool, a GitHub repo with a readable README, a personal site that does something. This is the single highest-leverage change you can make for a builder-disposition screen.
What belongs in Shipped:
- A Lovable or v0 app with a live URL, even if it has fifteen users.
- A GitHub repo you actually pushed to in the last 90 days, not a fork from 2021.
- An internal tool you built at work. Describe it; link if public. Dropbox's Nova is the archetype.
- A personal site that is not a template. Hand-rolled beats polished-template every time for this screen.
- An eval harness, a prompt library, or a small open-source agent. These read as judgment, not just output.
What does not belong: tutorials you finished, Coursera certificates, "contributor" badges from one-line PRs, hackathon submissions you never deployed. The screen is not fooled by activity theater. Hiring managers have seen the pattern enough times to spot it on sight.
For non-engineers, the bar is lower but the category is the same. A marketer with a Lovable landing-page generator, a PM with an n8n workflow that routes customer feedback, an ops lead with a Retool dashboard pulling from three APIs. Dropbox's pattern of non-technical staff shipping Lovable apps means hiring managers now assume these tools are accessible to anyone. Absence reads as unwillingness, not inability.
The work-trial era is already here
Foxglove and Rounds now screen with AI-permitted work trials of 30 minutes to a full week, and your resume's job is to get you to the trial, not to simulate it. Foxglove's Neder tells candidates "we expect you to use AI, and we will give you whatever AI tools you want," and the firm extended offers to 8 of 13 people who completed trials in the last 90 days. Rounds, led by CEO Fardeen Khimani, runs 30 to 90 minute AI-permitted simulations across SWE, design, marketing, content, and PM roles.
Candidates were encouraged to use any AI tools during the week-long trial.
Two implications for the resume stage:
- The resume no longer has to prove you can do the work. The trial does that. The resume has to prove you have the disposition.
- Tool fluency bullets that cannot be verified in a trial are a liability. If you claim daily Cursor use, the trial will reveal in twenty minutes whether that is true.
Write the resume with the trial in mind. Every tool you list, you should be able to open and use fluently in a shared screen. Every artifact you link, you should be able to walk through the architecture of on a call. If a bullet cannot survive a follow-up question, cut it.
How to tailor this for the posting in front of you
For each application, rewrite three things against the specific posting: your Shipped section (lead with the artifact closest to the company's stack), your top three bullets (judgment-shaped, with the company's AI tools named), and your summary line (replace "FAANG engineer" with "builder who ships with [their tools]"). This is per-posting work, not a one-time rewrite.
Artifacts don't travel equally in a builder-disposition world. A Supabase and Lovable app is gold for a Supabase-native shop and ambient for a Rails shop. A Cursor-indexed monorepo refactor matters at a 550k-file codebase like Dropbox's and reads as overkill at a 20k-file one. The ordering of your Shipped links should change by employer, and the artifact you lead with should match their stack.
If doing this three times a week sounds exhausting, that is what Refolk is built for. It drafts the tailored resume and cover letter from your own history against the posting, and scores how well your actual background fits before you spend an hour on an application you were never going to clear. You still own the artifacts. The rewrite is mechanical.
The last piece is honesty about fit. The 10x gap between daily AI users and AI-aware candidates is real, and it cuts both ways. If you have been reading about Cursor rather than using it, pick a weekend project before you apply. Ship something small. Put the link on the resume. That one act moves you from the losing side of the disposition screen to the winning side faster than any bullet rewrite.
FAQ
Should I remove my FAANG line entirely?
No. Keep it, but demote it. A FAANG line still signals you can pass a bar interview and work in a large codebase, and plenty of non-AI-native teams still weight it. Move it to a single line in your work history and let the Shipped section take the top of the page. Pedigree is not worthless; it has lost its position as the leading signal for a growing subset of employers, especially AI-native startups and mid-stage companies in the middle of an "AI-first" transformation like Dropbox's.
What if I cannot ship a side project because of my employer's IP agreement?
Build and ship something off-topic to the company. A board game scorer, a tool for your running club, a small CLI for a hobby. Builder's disposition does not care about domain, it cares that you make things. If even that is blocked, write up a public technical post about an eval you built internally, with the specifics anonymized. The signal is "I think like a builder," not "I shipped a competing product."
How do I show AI skills on my resume without keyword stuffing?
Replace the Skills row's AI section with three bullets inside your experience that each name one tool, one decision, and one measurable result. For example: "Chose Cursor over Copilot for a 180k-file migration after a one-week bakeoff; team ramped from 0 to daily use in two sprints." That one bullet signals Cursor fluency, judgment, and leadership, which is three jobs the keyword row did badly.
Does builder's disposition apply to PMs, designers, and marketers too?
Yes, and more so because the bar is lower. The Dropbox pattern of non-technical staff shipping Lovable apps means PMs, designers, and marketers are now expected to have at least one shipped thing: a prototype, a workflow, a scraper, a dashboard. The absence of any shipped artifact in a non-engineering resume used to be normal and is now a flag. Pick one tool (Lovable, v0, n8n, Retool), ship one small thing, put the link on the resume.