Is collaboration mode the same as the LockedIn Duo feature?

Conceptually similar; the mechanic differs in three ways. First, our helper channel is read-only on the candidate side — the helper can send text prompts but cannot take screen control or click anything. Second, audio is buffered on the helper side with about 200ms of latency, not synchronised to the millisecond, so the helper can react to a sentence without losing the next one. Third, the helper sees no persistent state from the candidate's account; what they see is the current session and nothing else.

Will the interviewer see anything that suggests a second person is on the call?

No. The helper's window is on a separate device entirely; nothing about it is captured by the candidate's meeting client. Audio and video on the interviewer's side show only the candidate. From a network-traffic perspective, the candidate's machine looks like a normal Acedly session. The honest disclosure caveat: if your employer's interview policy prohibits additional participants on a call, the responsibility for disclosing this stays with you regardless of detectability.

Can I use collaboration mode for a coding round?

Technically yes, in practice we don't recommend it for FAANG-tier coding panels. Top-tier recruiters are increasingly trained to spot mismatched typing rhythms; adding a second human in the loop introduces cadence noise on top of any solo-copilot assistance. The solo copilot handles coding rounds cleanly; collaboration mode adds the most value in system-design, founder, and senior-judgement rounds.

Can my helper see my résumé or my prior interview history?

No. The helper sees only the current session — the live transcript and a small input field for sending prompts. Your résumé, knowledge base, prior session recordings, and personal information stay private to your account. After the session ends, the helper retains nothing on their account beyond a session record they can delete.

What if I want to revoke the helper mid-interview?

One tap from the collaboration panel ends the helper's read-only window immediately. The interview itself continues uninterrupted on your side; the helper just falls off the channel. You can also pause the channel temporarily for any moment of the call without ending it — useful when the interviewer asks a question you want to answer entirely on your own.

Is using collaboration mode legal where I am?

The category sits in a legal grey zone that has been tightening since 2023. The Illinois AI Video Interview Act, NYC Local Law 144, the EU AI Act, and similar legislation regulate the *employer* side of automated screening; the *candidate* side is still mostly governed by general contract and employment law. The practical line is the employer's interview policy: if the policy explicitly prohibits additional participants on a call, do not use this mode in that interview. We surface the policy question on the product page deliberately so that candidates can make an informed choice rather than discover the issue post-hoc.

Who is the best helper for me to invite?

Someone who has done the round you're about to do, at the level you're targeting, in the recent enough past that the question patterns are still familiar to them. A friend at the company is ideal. A senior engineer two levels above you is the next best. A paid coach you trust is third. A friend who hasn't interviewed in three years is at best a moral-support helper, not a substantive one — invite them for the comfort, but don't expect strong prompts.

Product Guide12 min read

Remote Interview Collaboration: How a Trusted Helper Can Assist During a Live Call (2026)

What a remote-help interview workflow actually looks like, how Acedly AI's collaboration mode keeps the helper invisible to the interviewer, when this format makes sense, and the ethical line — from the team building it.

Devon Park

Head of Research, Acedly

Published May 15, 2026

What "collaboration" actually means in a live interview

The most accurate frame for this category is peer-review at conversation speed. A second person — usually someone the candidate knows well, occasionally a paid coach — joins a live interview session through a separate device on the helper's side. They see what the candidate sees and hear what the candidate hears, but in a read-only window that produces no outbound audio, video, or screen artifact. The candidate, on their side, sees the helper's text prompts on the same invisible overlay that displays the solo copilot's output.

The format exists because most candidates with a strong network already do an informal version of it: a chat group on the phone during a phone screen, a friend on the same call as a "silent observer," a mentor available by text between rounds. The collaboration mode formalises and de-risks this practice — single channel, audio synchronisation, OS-level stealth on the candidate's side, no separate device juggling.

It is not a replacement for the solo real-time copilot. A solo copilot is an LLM that drafts; a collaborator is a human who reads, judges, and writes back. The two are useful in different rounds for different reasons, and many candidates use them together.

How Acedly's collaboration mode works

The mechanic is intentionally narrow. The collaboration is a side channel between two endpoints — the candidate's machine running the live copilot, and a helper's machine running a read-only browser session. The helper joins through a one-time invite link generated from the candidate's side; nothing about the session persists on the helper's account beyond a session record they can delete.

Audio synchronisation. Both endpoints hear the interview audio at the same time, with a buffered latency of about 200ms on the helper's side. The buffering is there so the helper can react to a sentence in the interviewer's question without missing the next sentence; it does not introduce noticeable lag on the candidate's side.

Text-only outbound from helper. The helper can send short text prompts to the candidate's overlay — a one-line hint, a structural suggestion ("you skipped the constraint about read-heavy"), or a reminder of a story you've discussed before the interview. The prompts surface as a separate panel on the candidate's invisible overlay, distinguished visually from the solo copilot's drafts. There is no audio, no video, no shared cursor.

Stealth on the candidate side. The helper's prompts are subject to the same OS-level capture exclusion as the solo copilot — NSWindowSharingNone on macOS, WDA_EXCLUDEFROMCAPTURE on Windows. From the interviewer's screen-share, the prompts do not exist.

No persistent helper state. The helper does not see the candidate's résumé, knowledge base, or prior interview history. They see only what's happening in the current session. After the session ends, the helper retains only what they wrote down themselves.

Single helper, candidate-controlled. One helper per session, invited by the candidate, with a one-tap revoke that immediately ends the helper's read-only window. The candidate can also pause the channel for any moment of the call without ending it.

When collaboration helps and when it doesn't

The honest read on this category is more nuanced than the marketing pages of competitors that sell similar features tend to admit.

Where collaboration meaningfully helps:

Senior system-design rounds. A helper who has done the round before can flag the trade-off you're about to miss, the deep-dive component you're under-investing in, or the moment you're driving past instead of pausing to ask. The signal these rounds reward is judgement under ambiguity; a second pair of eyes catches misjudgements faster than a model.
Founder and hiring-manager rounds at startups. The question behind the question matters more than the question itself. A collaborator who knows the company can name the real concern the interviewer is probing — "she's trying to figure out whether you're a strong-opinions-loosely-held person; lean into the disagreement story."
Late-stage rounds with a high stakes mismatch. The 11th round of a 12-round senior loop. Final-round case interviews. Partner interviews at consulting firms. The marginal value of a second judgement here is large because the marginal cost of a misstep is large.
Non-native-language interviews where the candidate is strong on substance but weak on idiom. A helper fluent in the interview language can suggest the precise phrasing without writing the answer.

Where collaboration doesn't help (or hurts):

High-volume early screens. A first-round recruiter call is too short and too scripted for a collaboration channel to add value. The solo copilot handles it cleanly; adding a helper introduces a coordination tax for no return.
Pure coding rounds with strong typing-rhythm tells. Recruiters at top-tier companies are increasingly trained to spot mismatched typing rhythms. A helper sending coding hints introduces additional cadence noise on top of any solo copilot use. If you're going to use AI assistance in a coding round, use the solo copilot; do not chain a helper on top.
Rounds you would have passed on substance. Adding a collaborator to a round you'd pass anyway adds risk without return. The category is highest-leverage at the rounds where you'd be borderline.
Rounds you can't pass on substance. The mirror failure. A collaborator cannot substitute for years of preparation; the prompts will be at most one notch better than your own thinking. If the round is beyond your reach, more sessions and more drilling outperform more helpers.

The ethical line, and where this product draws it

The category is contested, and Acedly's stance is explicit rather than evasive.

Many employers treat real-time AI assistance the same way they treat note-taking or referencing your own résumé during a call: appropriate, expected for some candidates, not disclosed as a matter of course. Many of those same employers treat another person listening to the call as a different category — one where disclosure is more often expected and where the absence of disclosure is more often considered a misrepresentation.

This is not a position we want to soften. A collaborator is closer to "another person joined the call" than to "I had notes open." For the rounds where this matters — most professional rounds where the employer has a formal interview policy — the responsibility for disclosure stays with the candidate. Acedly's collaboration mode is built with that responsibility in mind: it does not pretend to be invisible on principle, and it does not hide the fact that a second human is on the channel from the candidate themselves.

The product also makes one design choice we want to be honest about: the prompts are short, text-only, and visibly separated from the solo copilot's output on the candidate's side. This is deliberate friction. It is designed to prevent the failure mode where a candidate types verbatim what a helper writes and gets caught on a follow-up question because the cadence is wrong and the substance is one inch off from what they actually understood. A real-time copilot is a thinking aid; a real-time collaborator is a cross-check, not a teleprompter.

The candidates who get the most value out of this mode are the ones who would have done well in the round without it. They use the helper to catch a missed constraint, a forgotten metric, the moment when a recruiter's body language changes — not to substitute for preparation they haven't done.

How a collaboration session feels in practice

The end-to-end of a collaboration-mode session is small enough to describe in a paragraph.

Before the interview, the candidate opens Acedly, picks the session type (technical / behavioural / system-design / case), and generates a one-time invite link from the collaboration panel. The link expires in fifteen minutes if unused. The candidate sends it to the helper through whatever channel they prefer.

During the interview, the helper sees a read-only browser window with the live transcript and a small input field for sending prompts. The candidate sees the helper's prompts in a distinct panel on the invisible overlay, separated visually from the solo copilot's drafts. Audio is synchronised; both endpoints hear the interviewer at roughly the same time.

After the interview, the session record is stored on the candidate's account only — the helper sees a closing summary on their side but has no persistent access. The candidate can review the session, including the prompts the helper sent, as part of the post-interview review flow.

Solo real-time copilot vs collaboration mode

Feature	Solo copilot	Collaboration mode
Who's on the channel	Model only	Model + one trusted helper
Best for	Volume rounds, structure recall, coding	Senior judgement rounds, founder rounds, non-native idiom
Setup time	None — open and go	1 minute to send invite
Coordination tax	None	Helper must be available at interview time
Marginal value over no AI	Large, especially at sub-200ms latency	Largest where judgement matters most
Risk profile	Same as note-taking in most policies	Higher — second human present