2026 Capital One AI Coding Interview (Agentic Round with Claude Code), New Grad TDP Power Day: 7 Experiences
- Company
- Capital One
- Reports
- 6 interview reports
- Level
- New Grad
- Stages
- Onsite
In the agentic round one TDP candidate described, you get about 20 minutes to fix a bug in a large app with Claude Code, then about 20 more to add a feature. Four of these seven new grads ran out of time, and two of those four still got offers.
Coverage: 7 new grad candidates in Capital One's Technology Development Program (TDP), Associate level, February 2027 start · The agentic coding round of the Power Day, with short notes on the case and behavioral rounds where candidates mentioned them · Interviews from about September 30 to October 23, 2026 · One location given (Dallas, chosen at the offer stage)
What came up
Every one of the seven reports describes Capital One's AI coding interview, the agentic round, as two separately timed parts. Most candidates wrote about how far they got rather than what the task was. Four ran out of time before finishing both parts, and two finished early.
| Topic | Reports | Most recent report |
|---|---|---|
| Agentic round in two timed parts | 7 of 7 | Oct 23, 2026 (interview) |
| Ran out of time before finishing both parts | 4 of 7 | Oct 23, 2026 (interview) |
| Feature-building part named | 4 of 7 | Oct 23, 2026 (interview) |
| Finished both parts | 2 of 7 | October 2026 (interview; posted Oct 25) |
| Bug-fix part named | 3 of 7 | Oct 7, 2026 (interview) |
| Explaining their reasoning to the interviewer | 2 of 7 | Oct 7, 2026 (interview) |
| Large full-stack app in CodeSignal, with Claude Code | 1 of 7 | Sep 30 – Oct 4, 2026 (interview, inferred) |
These counts cover all 7 reports, and one candidate can appear in several rows. None of the seven described the actual bug or feature they were given.
Questions we found
AI Coding
| Question | Frequency | All last seen dates |
|---|---|---|
| After a code reset, add a new feature to the same large full-stack app | 4 reports | Oct 23, Oct 8, Oct 5, 2026; Sep 30 – Oct 4, 2026 (interview dates) |
| Find and fix a bug in a large full-stack app | 3 reports | Oct 7, Oct 5, 2026; Sep 30 – Oct 4, 2026 (interview dates) |
Reports 02 and 05 went through both parts without saying which was which, so they aren't counted in these two rows.
The Capital One agentic round, as one candidate laid it out
This section keeps one candidate's account whole. It is the only one of the seven that walks through how the round works; the six timelines after it are shorter and mostly say how far each person got.
Agentic 01: A codebase too big to read, two 20-minute tasks, and a reset in between
Interview date: between September 30 and October 4, 2026 (inferred from the candidate's earlier posts) · Report date: October 10, 2026 · Role: TDP Associate, February 2027 start · Result: Recruiter call booked as of October 24; outcome not posted
Before the Power Day
- CodeSignal General Coding Assessment: a full 600, submitted about ten minutes early.
- A logistics email asking about things like degree and visa status arrived on September 23. Interview scheduling came after it.
Agentic coding
- The round ran in CodeSignal's web IDE. Inside was a large full-stack application. The candidate could open as many terminals as they liked and use Claude Code without limit, though a few of its settings had been changed.
- Part one: about 20 minutes to find and fix a bug, then about 5 minutes talking it over with the interviewer.
- The code was reset.
- Part two: about 20 minutes to build a new feature in the same app, then another 5 minutes of discussion.
- The codebase was far too large to take in quickly. Looking back, the candidate said you clearly weren't expected to type code yourself unless the agent got something wrong.
- They found the round tough. They also pointed out that most agentic-interview prep they could find copies Meta's version, which has a smaller and simpler codebase but leans less on the agent and more on data structures and algorithms.
If they were preparing again, they said, they would pick an open-source repo too big to understand at a glance and ask Claude to plant a bug, describe how the bug shows up in use, and write a short spec for a new feature. Then they would work each one in 20 minutes, resetting everything in between, and practice thinking out loud so it's clear they're doing more than letting Claude run.
Technical case
- It closely followed the roughly 30-minute example video Capital One provides: a business discussion about a product, then tracing a set of business rules by hand, then debugging the code that applies those rules, then writing a new method that does something similar.
- No code could be run, so anything written had to be checked by eye. The candidate called the round very easy to slip up in, but fine for a clear communicator who can program.
Behavioral
- About as straightforward as interviews get. They called two prompts very common, based on their own Power Day, other candidates and their prep, without saying which ones they were asked: a time you challenged the status quo, and questions about how you use AI (practice).
Afterward
- The candidate talked with dozens of others who had offers, rejections or were still waiting. Few felt they had nailed the whole day, and only one described the agentic round as going well.
- Their own wait ran long. By October 24 they had a recruiter call on the calendar. From what they had heard, a scheduled call almost always meant an offer, while rejections arrived by email.
6 detailed candidate timelines, loops, and screens
01: A perfect case study, then stuck in part two
Interview date: October 23, 2026 · Report date: October 24, 2026 · Role: TDP, February 2027 cohort · Result: Still waiting when they posted
Power Day
- Technical case: the candidate rated it perfect.
- Behavioral: strong, in their view.
- Agentic coding: in the second part they couldn't get their implementation debugged before time ran out. They posted mainly to ask whether they still had a chance.
02: The agentic round was the easy one
Interview date: October 2026, about eight business days before they heard back · Report date: October 25, 2026 · Role: TDP, February 2027 cohort · Result: Recruiter call booked; outcome not posted
Power Day
- Technical case: the business analysis and debugging went quickly. Tracing the rules and writing the function took some fumbling, but they got there. That part left them worried about the result.
- Behavioral: standard, and over fast.
- Agentic coding: easily their best round. They finished both parts in about half the time allowed.
They shared this partly because they hadn't realized others were struggling with the agentic part. The report doesn't say how much code they wrote by hand.
03: Both tasks done, rejected the next day
Interview date: October 8, 2026 · Report date: October 9, 2026 · Role: TDP, February 2027 cohort · Result: Rejected on October 9
Power Day
- Technical case and behavioral: they felt both went very well, with strong feedback from each interviewer.
- Agentic coding: they finished without trouble and had the new feature working within the time. Technically, they rated this round very strong.
- They got along with all the interviewers, but thought the last one, from the agentic round, seemed tired and ready to be done. They couldn't tell whether that mattered.
The rejection came a day later, after what they had considered one of their best interview days.
04: Only the bug fix, and an offer two days later
Interview date: October 7, 2026 · Report date: October 11, 2026 · Role: TDP, February 2027 cohort · Location: Dallas · Result: Offer on October 10
Power Day
- Technical case: fine overall, though they stumbled on the business assessment and settled on an answer that was good enough.
- Behavioral: fine, with good rapport.
- Agentic coding: their weakest round. They solved only the debugging task, and made sure to explain their thinking as they worked.
After the Power Day
- The next morning an email asked them to book a call with their recruiter. On that call, October 10, the recruiter told them they were getting an offer.
- For a February start, the only locations available were Dallas and a Virginia site they believed was Richmond. They chose Dallas.
At the time they had put their own chances at about even.
05: Didn't finish both tasks, offer on October 16
Interview date: October 6, 2026 · Report date: October 9, 2026 · Role: TDP, February 2027 cohort · Result: Offer on October 16
Power Day
- Technical case: they made some mistakes but caught and corrected them.
- Agentic coding: they couldn't get through both tasks in time and weren't sure how much that would count against them.
Overall they called the day fine. The offer came ten days after the interview.
06: Bug fixed, feature not reached
Interview date: October 5, 2026 · Report date: October 9, 2026 · Role: TDP, February 2027 cohort · Result: No decision as of October 9
Power Day
- Technical case and behavioral: decent, by their account.
- Agentic coding: they fixed the bug, then ran out of time before building the feature.
Patterns across the reports
-
Finishing the agentic tasks didn't decide the outcome. Four candidates (01, 04, 05 and 06) didn't complete both parts, and two of them, 04 and 05, got offers. Report 03 finished both parts and was rejected the next day. Report 02 finished in half the time and had a recruiter call booked. The candidates also disagreed on how hard the round was: 02 called it their best round, 04 their worst, and the candidate in Agentic 01 called it tough.
-
The second part is where time ran short. In 06 the bug got fixed and the feature never started. In 01 the trouble came while debugging the second part's implementation. Report 04 also finished only the debugging task.
-
Talking through the work came up more than once. The candidate in Agentic 01 describes a five-minute discussion after each part, and would practice thinking out loud to show they were doing more than letting Claude run. In 04 the candidate solved only one task but focused on explaining their thinking, and got an offer. Neither report says how the interviewers weighed that.
Keep practicing
- Tell me about a time you challenged the status quo: a prompt Agentic 01 called very common.
- What are your thoughts on AI?: the closest Capital One practice prompt to the "how do you use AI" questions in Agentic 01.
- Capital One project coding questions: Capital One's codebase-style practice projects. Sign-in required.
- Capital One behavioral questions: every Capital One behavioral prompt on FastPrep. Sign-in required.
- 2027-Cycle Capital One TDP New Grad OA / Power Day: 24 Experiences: the whole TDP process, with the OA, the technical case and behavioral rounds, and summer 2026 Power Days that still had a regular coding round.
- 2027-Cycle Capital One TIP Intern OA / Power Day: 12 Experiences: the intern cohort, including an October 2026 intern Power Day with a Claude Code round.
- Capital One interview guide: Capital One's interview process and question list.
- Capital One hiring insights: hiring data for Capital One.
- Experience feed: candidate reports from every company.
AI coding rounds at other companies
- Meta: Meta AI coding interview: two onsite rounds where you finish a hard algorithm inside a mostly written codebase with an AI model, then face stress-test suites.
- Google: Google AI coding interview: an online assessment with HackerRank's built-in assistant, and a live round on a multi-file project where using AI was optional.
- Stripe: Stripe AI coding interview: experienced engineers kept getting the same transaction rule parser, built step by step with a built-in model.
- Stripe interns: Stripe intern AI coding round: ten intern accounts of the same AI-assisted round.
- Rippling: Rippling AI coding interview: HackerRank's built-in AI on Rippling's familiar problems, plus an HR codebase at the onsite.
- DoorDash: DoorDash AI coding interview: AI CodeCraft: build and run several small services live with your own AI tool.
- LinkedIn: LinkedIn AI coding interview: one long, multi-part task in an IDE whose agent edits files directly.
- Shopify: Shopify AI coding interview: pair programming with your own AI on Shopify's well-known question set.
- Datadog: Datadog AI coding interview: bring any AI tool and share your screen; the tasks are easy, and the follow-ups decide the round.
- Bloomberg: Bloomberg AI coding interview: intern rounds that start in a small file and move to a large codebase, debugging with AI help.
- xAI: xAI AI coding interview: new grad take-homes and onsites where you build with AI.
- Adobe: Adobe AI coding interview: AI showed up as one step in a director round and in a take-home; the rest stayed classic.
- Snowflake: Snowflake AI coding interview: design and build a rate limiter or a retrying task queue while the AI writes the code; one AI Engineer onsite was in person with your own laptop.
- Snap: Snap AI coding interview: a backend-only chat service from an empty Python project with Claude or Codex, or a feature in an existing codebase.
- Ramp: Ramp AI coding interview: build inside a repo Ramp provides; one repo logged every prompt for the interviewer to read.
- Anthropic: Anthropic AI coding interview: review four pull requests, then improve one with Claude Code.
- Apple: Apple AI coding interview: CoderPad with a sidebar agent, alongside Apple's other recent interviews.
More companies with AI rounds on FastPrep:
- Amazon: Amazon interview report: an AI-assisted debugging OA (1 experience).
- Atlassian: Atlassian interview report: project coding and AI tool use (11 experiences, two parts).
- Circle: Circle interview report: three AI rounds: building, debugging and PR review (1 experience).
- Volta: Volta interview report: a deliberately vague data task with a token budget (1 experience).
Across companies: 2026 AI Coding OAs at 15 Companies: 32 Reported Problems and 2025–2026 AI Coding Phone Screens and Onsites at 43 Companies: 76 Reported Problems.
Sources: Anywhere useful interview info can be found.
Comments
0No comments yet — be the first to share what you know.