What 5 Onsite Loops at 5 Different Companies Actually Looked Like in 2026 — Hour by Hour

I have been through 5 onsite loops in the last 90 days. Different companies, different stages, different formats. Three offers, two rejections. I took notes inside each loop — what came when, who I met, what they asked, how long it ran.

This post is the comparison. If you are about to walk into your first senior onsite of 2026, this is what the variance actually looks like.

The five loops

I'll describe the loops by stage, not by name (companies asked).

• Loop A — FAANG-scale public company, senior backend role
• Loop B — Mid-cap public, infrastructure team
• Loop C — Late-stage private (~2000 engineers), platform role
• Loop D — Series C startup (~150 engineers), staff-track backend
• Loop E — Series A startup (~30 engineers), tech-lead role

Loop A — FAANG, 6 hours

A single Friday, 9am-3pm. Five interviews back-to-back, 50 minutes each, with 10-min breaks. Lunch was solo at the desk — not optional, just nobody scheduled time for it.

Sessions:

1. Coding (medium-hard graph problem, 2-language choice, whiteboard-coded)
2. System design (a real-time messaging system at scale, no specific scale numbers given)
3. Behavioral / leadership (STAR-format, 6 questions in 50 mins)
4. Domain-specific (database internals, mostly conceptual)
5. Bar-raiser (mixed coding + behavioral, with a senior staff engineer not on the team)

What predicted the offer (didn't get it): I did fine on coding, fine on system design, was lukewarm on the bar-raiser. The bar-raiser is the gating signal for FAANG. The bar-raiser is not a normal interview. They are looking for one specific kind of pattern. Find friends in your network who have been through one and ask what their bar-raiser focused on.

Loop B — Mid-cap public, 4 hours

Two days, two hours each. Day 1: coding + behavioral. Day 2: system design + cross-functional (a session with a PM and a design engineer).

What was different from FAANG: the cross-functional session was a real signal. They wanted to see how I handled product disagreement, not just engineering trade-offs. I got the offer. The cross-functional session was the differentiator.

What predicted the offer: I had gone in expecting another behavioral round and instead got a soft product debate. I leaned in. The PM disagreed with my technical instinct, I said "okay, let me work through your version and see if I get the same conclusion." We worked through it for 20 minutes. By the end I was about 80% her version and 20% mine. They told me afterward that "willing to update on evidence" was the box they were trying to check.

Loop C — Late-stage private, 5 hours

One day, 9am-2pm. Four interviews + a hiring manager debrief at the end.

Format that surprised me: the first round was a 60-minute "live debugging session" against an open-source repo I'd never seen. I had two minutes to read the README and the failing test, then 58 minutes to fix it on a shared screen with the interviewer. The bug was real (it was in the git history). The interviewer's job was to evaluate my live-debugging process.

What predicted the offer (got it): I asked questions instead of guessing. Specifically, three of them: "what does this function do at runtime," "is the test the source of truth or is the spec," and "can I read the test fixture before guessing." All three were the questions the interviewer was scoring for. It is far more pleasant than whiteboard coding once you adapt.

Loop D — Series C, 4 hours over a week

Three sessions across a week. 90 minutes each. No "loop day." The interviews were spread out and felt much more like ongoing conversations than a high-pressure event.

Sessions:

1. Take-home walkthrough (I had submitted a take-home a week prior; the session was a 90-min discussion of my code)
2. System design (specific to their actual product, not a hypothetical)
3. Hiring manager + cofounder + a senior engineer

What predicted the offer (got it): the take-home walkthrough is where the loop is won. They had read the code. They had specific questions. I had to defend choices I had made a week earlier. Spend more time on the take-home than you think you should. The bar is "could you maintain this in 6 months yourself." Mine was barely above that bar.

Loop E — Series A, 5 hours but informal

One day, but felt nothing like a loop. Office, espresso machine, four 60-minute sessions including the founder. The technical content was genuinely the easiest of the five loops. The fit content was the hardest.

The founder asked me variants of the same question for 45 minutes: "if you joined and the first month went badly, what would you do?" Bad month after bad month. The pressure was not technical. It was "do you have the temperament for an early-stage company."

What predicted the rejection (didn't get it): I gave clean technical answers. I gave okay-but-not-great founder-fit answers. They went with someone who was technically less strong but who they liked more. At Series A, fit wins over skill. The reverse is true at scale. The earlier you go, the more this matters.

What was the same across all five

Three patterns held across every loop, regardless of stage:

1. The interviewers had read the resume. Every interviewer (across 19 total interviewers) had at least skimmed it. None of them asked me to recap my background. This is the new normal.
2. System design questions were grounded in the company's real product. Generic "design Twitter" was the exception, not the rule. They wanted to see whether I could think about a problem they actually had.
3. The hiring manager's debrief was the highest-stakes interview at every stage. This is also where I made the most mistakes early. The debrief is not a casual chat. It is the synthesis interview. Treat it accordingly.

What changed by stage

FAANG: rigorous + impersonal. Optimize for "exact technical answer." The bar is technical depth.

Mid-cap public: technical + cross-functional. Optimize for "willing to update on evidence." The bar is collaboration.

Late-stage private: practical + applied. Optimize for "can navigate ambiguity." The bar is applied judgment.

Series C: code-first + take-home-heavy. Optimize for "code I can defend." The bar is craft.

Series A: founder-fit-first. Optimize for "would survive bad months." The bar is temperament.

Different stages are not the same interview with a different vibe. They are different interviews. Tailor your prep to the stage you are interviewing at.

What I'd change about my own prep

Two things, in retrospect.

1. I over-prepared for system design and under-prepared for hiring-manager debriefs. The debrief decided 4 of 5 outcomes. System design tied or moved the needle on 2 of 5. The energy budget was wrong.

2. I didn't research the interviewer's background before each session. Three of the times I tanked an interview, the interviewer had a specific background I would have been able to find on LinkedIn in 90 seconds. Knowing in advance that the interviewer was a former search infrastructure engineer would have changed how I framed an answer.

For the next loop I run, debrief prep + interviewer research come first. System design prep comes second. Coding stays where it is.

Free tools I built for the job search: resume-checker, job-keywords, resume-bullets. All free, all in the browser, no signup.

Earlier in this series:

• I Lost an Onsite Loop and Got the Real Debrief — the postmortem
• The Week Before the Onsite: A Day-by-Day Plan — pre-loop prep
• 8 Anti-Patterns That Kill Take-Home Code Submissions — take-home angle