Bloomberg SWE Intern AI-Assisted Coding Round: 7 Experiences
- Company
- Bloomberg LP
- Reports
- 7 interview reports
- Level
- Internship
- Stages
- Onsite
Bloomberg's fall 2026 SWE intern loops ran AI-assisted codebase rounds: fix a small seat-booking program, check code against a spec, or work in a big repo.
At a glance
- Process: two loops had two AI-assisted codebase rounds, a small file or codebase first and a large one second (entries 02, 05); in two loops a system design round came after the codebase work (entries 03, 05).
- Asked recently: a bug hunt in an existing codebase, where the agent's failing tests showed where the bugs were (1 of 7, entry 01, the newest report); fixing the bugs in a short seat-booking or seat-allocation program (2 of 7, Python in entry 04); finding where code breaks against a spec or an issue (2 of 7); four tickets in a large matchmaking system (1 of 7).
- Difficulty: entry 01 rated its round easy and still didn't get an offer; entry 03 called its ticket round hard and still passed with 2 of 4 tickets; entry 06 fixed one bug with a hint before time ran out; five of the seven did not get an offer.
- Practise first: Debug Seat Hold, Confirm and Cancel Rules for the seat-booking debugging in entries 04 and 06, and Resolve Tickets in Multi-Tenant Matchmaking for entry 03's ticket round.
Coverage: 7 software engineering intern candidates (two in London, one in Warsaw, four with no location given, one recruited on campus) · AI-assisted coding and codebase rounds, plus the system design round that followed in two loops · interviews October–November 2026
Other levels: Bloomberg new grad AI-assisted coding rounds, Bloomberg experienced engineer AI coding interviews and Bloomberg AI interview reports with no level given. Every level in one place: Bloomberg AI-assisted coding round, all levels.
What came up
Six of the seven reports describe work inside code the candidate had never seen. The two with the clearest tasks were both a short program for handing out seats, and the newest was a bug hunt led by tests the agent wrote.
| Topic | Reports | Last seen |
|---|---|---|
| Tasks inside existing, unfamiliar code | 6 of 7 | Nov 2026 (report date) |
| Finding where code breaks and fixing it | 5 of 7 | Nov 2026 (report date) |
| Interviewers watching how the AI assistant is used | 3 of 7 | Oct 2026 |
| A small file or codebase first, then a large one | 2 of 7 | Nov 2026 (report date) |
| Seat booking or seat allocation code with bugs | 2 of 7 | Oct 2026 |
| Unclear or incomplete requirements to clarify | 2 of 7 | Oct 2026 |
| Failing tests from the agent to locate the bugs | 1 of 7 | Nov 2026 (report date) |
| Ticket-style tasks in a large system | 1 of 7 | Nov 2026 (report date) |
| Testing the fix and explaining it | 1 of 7 | Oct 2026 |
Questions we found
AI Coding
| Question | Reports | Last seen |
|---|---|---|
| Fix the bugs in a short seat-booking or seat-allocation program | 2 | Oct 2026 |
| Locate bugs in an existing codebase through agent-generated failing tests, then repair them | 1 | Nov 2026 (report date) |
| Find the spec's ambiguities and the code that doesn't follow it, small codebase then large | 1 | Oct 2026 |
| Root-cause an issue in an unfamiliar repository, fix it, test it and explain the AI's part | 1 | Oct 2026 |
Project Coding
This report doesn't say whether an AI assistant was part of the round.
| Question | Reports | Last seen |
|---|---|---|
| Resolve four tickets in a large matchmaking system with queues and workers | 1 | Nov 2026 (report date) |
The system design prompts in entries 03 and 05 weren't shared, so they have no row.
7 detailed candidate timelines
01: Failing tests from the agent point to the bugs
Report date: Nov 18, 2026 · Role: Software Engineer Intern · Applied: Online · Result: No offer (the candidate still rated the experience positive)
Round 1: bugs in an existing codebase, with an AI agent
- The candidate had an AI agent to work with and an existing codebase that already had bugs in it.
- The bug hunt went test-first. The agent produced tests that fail where the code is wrong, and the code was repaired after that.
- They rated the round easy and found the interviewer nice, but no offer came. The report names no language, agent or time limit.
- This may be a later report from the candidate in entry 02, since nothing recorded tells the two apart. It is still counted: the recent intern reviews that could repeat earlier entries can't all be repeats, so at least one is a new person, and this one gives the most detail about the task.
02: One small file now, a large codebase next
Report date: Nov 8, 2026 · Role: SWE Intern · Result: In process (round 2 not yet taken)
Round 1: AI-assisted, single file
- The round used one small file, which the candidate found manageable. They didn't share the task.
Round 2 (scheduled): AI-assisted, large codebase
- They had been told round 2 would be AI-assisted inside a large codebase.
- Their worry was the jump in size: how to find their way around a big codebase, and how exactly the AI assistant was meant to be used.
03: Four matchmaking tickets, then an open-ended design
Report date: Nov 8, 2026 · Role: SWE Intern · Recruited: On campus · Result: Rejected after round 3
Round 1 (on campus)
- No content shared. The interviewer asked for availability for the next round right as it ended.
Round 2: tickets in a large codebase
- Four tickets in a huge codebase for a matchmaking system, full of queues, workers and business rules.
- The candidate found it hard; the interviewer was relaxed.
- They closed one ticket in about 30 minutes and another in about 10. Two of four was enough to pass, and the round 3 invite came about three days later.
- The report never mentions an AI assistant in this round. It sits here because it matches the large-codebase second round that entries 02 and 05 describe as AI-assisted.
Round 3: system design
- About ten minutes of resume talk, with one or two light questions.
- Then a very open-ended design problem, which they didn't share. The expected order was clarifying questions, API endpoints, a SQL or NoSQL choice with a schema, then a first high-level design, and time for their own questions at the end.
- Clarifying took longer than it should have, and the interviewer kept pushing them to start. Some early choices were wrong, so the design changed several times before they were happy with it.
- The rejection came about six days later. Looking for the problem afterwards, they found an easier version of it on a practice site, labelled hard there.
- Their preparation advice was lots of mock interviews. They practised with an AI in voice mode, but had little time to get ready.
04: A seat-booking class, an unclear spec and a project manager
Report date: Oct 30, 2026 · Interview date: October 2026 · Role: Software Engineer Intern · Location: London · Result: Accepted an offer
AI-assisted coding round
- One Python file of about 100 lines with a
Bookingclass that handles seats in three sections: VIP, Standard and Economy. The task was to find and fix its booking bugs. - The rules the code had to follow: holding seats lowers what is available; confirming a held booking must not take the seats a second time; cancelling a hold or a confirmed booking gives the seats back.
- The spec was left incomplete or vague on purpose, so part of the job was asking the right questions. The interviewer played a project manager.
- An AI assistant was built in. The candidate understood it was there to help with finding their way around the code, understanding it, debugging and making changes.
- What mattered, in their account: understanding the code, clarifying requirements, communicating, getting the fixes right, and testing edge cases.
05: A spec to check the code against, twice
Interview date: Oct 29, 2026 · Role: Software Engineer Intern · Location: London · Applied: Through the university · Process: About two weeks, three rounds · Result: No offer
Round 1: AI-assisted codebase interview (small)
- A small codebase and a written spec. The candidate had to spot where the spec was unclear, then find and debug the parts of the code that didn't follow it.
Round 2: AI-assisted codebase interview (large)
- The same kind of task in a large codebase.
- In both rounds the interviewers wanted to see the AI used for three jobs: clearing up the spec's unclear points, finding where code and spec disagree, and moving around the codebase.
Round 3: system design
- No details shared.
06: Two green tests and a seat-allocation bug hunt
Interview date: Oct 27, 2026 · Role: Intern (exact title not stated) · Result: Not successful (said in a later reply)
Round 1: AI-assisted debugging
- A program for allocating seats, about 100 to 150 lines, with a very short README. The two tests that came with it passed, yet bugs were still hiding in the code, and the job was to track them down and fix them.
- The assistant could switch into an agent that changed files on its own. The candidate leaned on it less than they could have, worried that handing it too much would count against them.
- About 30 minutes in, the interviewer offered a hint, and the two of them fixed one bug together before time ran out.
- Looking back, the candidate felt the interviewer cared more about getting to a fix than about the narration.
07: An issue in a repo nobody had shown them
Interview date: Oct 6, 2026 · Role: SWE Intern · Location: Warsaw · Result: No offer (the candidate still rated the experience positive)
AI-assisted codebase task
- Something was wrong in a repository the candidate had never worked in. With an AI assistant, they tracked down the root cause, wrote a fix and tested it.
- They also had to walk the interviewer through their reasoning and their approach, and explain how the assistant helped them reach the fix and check it.
Three more reports, possibly repeats
A software engineer intern review posted on Nov 18, 2026 describes a loop of five rounds: three technical ones (two AI-assisted coding interviews, then system design) and two behavioral ones, whose place in the loop isn't given.
- The second AI-assisted round used a much bigger codebase than the first.
- The candidate accepted an offer. They didn't share the tasks, the language or the tool.
Entry 04 also accepted an offer, and nothing in either review tells the two apart, so this may be the same person. It isn't counted in this article's totals.
An intern review posted on Nov 11, 2026 (interview in November 2026, applied through the university) describes a first round with AI help.
- The code was a single source file paired with its tests. The candidate had to repair it first, then extend it.
- An easy chat about their resume was part of the same round. They rated the difficulty average, didn't get an offer, and didn't name the tool or how they used it.
Entry 02's first round was also a single file, and entry 03 was also recruited through campus, ended without an offer and never shared its first round. Nothing in this review sets it apart from either, so it may be one of those two candidates. It isn't counted.
An intern review posted on Nov 17, 2026, with no fuller job title, mentions a first AI-assisted round.
- The candidate found the problem easy, and was surprised that nothing limited what the AI assistant would do.
- They didn't get an offer, and didn't share the task, the language or the tool.
Nothing in this review sets it apart from entries 02, 03 or 06, whose first rounds it could also describe, so it may be one of those three candidates. It isn't counted either.
Patterns across the reports
- Two AI-assisted rounds, small then large. Entries 02 and 05 both describe a first AI-assisted round on a small file or codebase and a second one in a large codebase. Entry 05 did both; entry 02 had only been told what round 2 would be.
- A short seat program with bugs. Entries 04 and 06 both got a program of 100 to 150 lines for booking or allocating seats (Python in entry 04). In 04 the spec itself was vague; in 06 the given tests passed even though bugs remained. One candidate got an offer and the other did not.
- The AI use is part of what is judged. Entries 04, 05 and 07 say the interviewers cared how the assistant was used: for navigating and debugging in 04, for reading the spec and the code in 05, and for reaching and checking the fix in 07. In entry 06 the candidate held back on the agent for fear of over-using it.
- System design after the codebase rounds. Entries 03 and 05 had a system design round after the codebase work, and both were rejected after it.
Common questions
What is Bloomberg's AI-assisted codebase round for interns?
You get code you didn't write and an AI assistant: a codebase whose bugs you surface with failing tests from the agent (entry 01), a short seat program with bugs (entries 04, 06), a spec to check the code against (entry 05) or an issue to root-cause (entry 07). Entry 03's large-codebase round was four tickets, though that report doesn't mention the assistant.
How many AI-assisted rounds do Bloomberg interns get?
Entries 02 and 05 describe two, the second in a larger codebase, and so does the uncounted Nov 18 review. Entry 03 had one large-codebase ticket round after an on-campus first round, without saying whether AI was part of it; entries 04, 06 and 07 each described one round, and entry 01 described only its first.
Do you need to finish every bug or ticket in Bloomberg's codebase round?
Not always. Entry 03 passed its ticket round with two of four tickets done; entry 06 fixed one bug with the interviewer's help and did not get through.
How did Bloomberg interns use the AI in the AI-assisted round?
In entry 01 the agent produced failing tests that pinned the bugs down, and the fixes came after; entry 04 understood it was there for navigating, understanding, debugging and editing; entry 07 had to explain how it helped reach and check the fix; entry 06 used the agent sparingly for fear that leaning on it would count against them.
Is Bloomberg's new grad AI round the same as the intern one?
It looks close. New grads also report a booking-program debugging round and a four-ticket matchmaking round; see the Bloomberg new grad AI-assisted coding rounds.
What to practise
- Debug Seat Hold, Confirm and Cancel Rules. A short Python booking class with seat holds, confirms and cancels in three sections, an AI assistant, and two passing tests that miss bugs. Built from entries 04 and 06 (2 of the 7 intern reports).
- Resolve Tickets in Multi-Tenant Matchmaking. Four tickets in a matchmaking engine with waiting queues and arenas, the kind of large-codebase round entry 03 had (1 of 7).
Keep practicing
- Debug Seat Hold, Confirm and Cancel Rules: the Bloomberg seat-booking debugging project built from entries 04 and 06.
- Resolve Tickets in Multi-Tenant Matchmaking: the matchmaking ticket project behind entry 03's round.
- Bloomberg project coding bank: every Bloomberg codebase exercise on FastPrep (sign-in needed).
- Bloomberg system design bank: for the design round in entries 03 and 05 (sign-in needed).
- Bloomberg interview guide: the company page with the full question list.
- Bloomberg hiring insights: recent interview activity at Bloomberg.
- Experience feed: more firsthand interview reports.
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.
- Capital One: Capital One AI coding interview: new grad Power Day agentic round: fix a bug, then add a feature to a large full-stack app with Claude Code.
- 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.