Over the last couple of issues, I’ve been talking about how companies are using AI-assisted rounds in different ways. This week, we have more intel about what to expect with AI-assisted rounds (and how companies are using them to hand out offers). 

Now, two more candidates are both saying they didn't finish their AI-assisted problem and still got offers. One reported a TIP offer and said they didn't finish either problem, but showed they knew how to use Claude Code responsibly. They used it to navigate the codebase, came up with their own hypothesis about what was going on, and then used it to validate that hypothesis, like writing a unit test to recreate the bug. As they talked through the solution, their interviewer kept asking why they prompted it that way and to identify the purpose of the prompt.

The other is a Capital One TDP offer. That candidate also didn't fully finish the assignment, and their read was that Capital One seemed to care more about the process and thinking behind their decisions than the completion itself.

These reports line up with what we heard about Microsoft's SWE round, and with the Figma advice we passed along earlier: show you're the one driving the AI, and you know how to “retain judgment and ownership of the result.” 

More intel on format. Like we said last week, the format of AI-assisted rounds still varies a lot (if the interviewer even uses one). Candidates are passing around an unconfirmed write-up that says Google replaced a traditional coding round with a Code Comprehension round in a May 2026 pilot. You'd get a 200-500 line unfamiliar codebase in CoderPad and 60 minutes to find bugs, discuss design decisions, and suggest improvements with Gemini as your AI. A candidate also asked me whether this applies to the Google SWE intern process, since their recruiter said it would be DS&A. The intern reports I’ve seen have all been DSA, and Google’s new grad looks like the pilot case for AI fluency.

Even within Microsoft, I’m getting different reports. A candidate who just got an offer for the AI/ML intern role described two 45-minute interviews: an LC medium focused on ML topics and an extensive ML system design round, with some behavioral at the start of both. They didn't mention an AI-assisted round like the AI SWE intern report. That's a different role title from the AI SWE intern report, and it’s not clear whether the gap is the team, the role, or just how it was reported.

If you want more information on AI-assisted rounds, check out last week’s intel here.

One candidate said their interviewer asked about orchestrating subagents and what they're for. They answered with the Bun Zig/Rust rewrite, where agents reviewed and wrote code, and got into agentic safety, like what agents should and shouldn't have access to. They also said you need good tech sense to get through these conversations.

A candidate who just finished the recruiter round said it was a resume walkthrough. They think the next stages are a coding round, a take-home assessment, and an onsite, and that the expectation is to be close to or as technical as a Stripe software engineer. It's one report, but I hadn't heard anything else on this process yet.

One candidate described a hiring screen/behavioral and a week-long PRD assignment where you propose an improvement to Databricks' product. The superday is three rounds: product sense, behavioral, and a PRD case study. They weren't asked explicit coding or SQL questions, but they did have to go deep on the technical side of past internship projects, including the stack and why they made certain decisions.

The Summer 2027 SWE intern email says they're evaluating candidates on a rolling basis and links a form for anyone with competing deadlines or offers. It matches what some candidates were told on campus. If you have a deadline coming up, fill it out, so your process can be expedited.

Meta is hosting an "Inside Meta Data Centers" info session and panel on Wednesday, October 28 at 6-7 pm EST. Register for this event to learn more about what the job actually looks like before a 12-week paid internship that will take place from June 7 - August 27, 2027. As I’ve shared before, companies are recording attendance at events like this to help make hiring decisions.

Register for the session →

OpenAI is hosting a "Work at the Frontier: Forward Deployed Engineering at OpenAI" livestream today on October 22, 2026, from 1:15-2:00 pm EST. 

Register for the event →

Rivian's Emerging Talent team is running its annual "Plug-in" virtual event series for Summer 2027 internships tomorrow, October 23, at 3:00 pm EST. The team will walk through its 2027 internship opportunities. Also, keep an eye out for positions opening in March or April.

RSVP for Plug-in →

If you've had an AI-assisted round this cycle, reply with the company and role, what setup they gave you (a codebase to debug, building from scratch, etc.), whether you finished, and whether the interviewer asked about your prompts or process. If you've heard back, tell us that too.

Subscribe to Direct Consideration to get the urgent roles, recruiter intel, events, and opportunities worth knowing about as they happen.