A.O.B.

Field notes

Shorter items, filed here rather than given pages of their own — because a page should exist when there's a page's worth to say, and no sooner.

Why the room always blames the remote folks

There is a structural unfairness built into every hybrid meeting, and once you see it you can't stop seeing it. The room cannot hear its own echo.

When a loop forms, it starts at a speaker in the room and leaves through a microphone in the same room, and the delayed copy is delivered to the far end. The table hears nothing wrong. The remote participants hear themselves half a second late, like a stadium announcer, and have to interrupt a meeting to report it — which is socially expensive, so they usually wait, and wait, and then say something apologetic at minute nine.

Meanwhile the room, hearing a perfectly normal meeting, concludes the problem is at the other end. "Your connection must be bad." "Try turning your camera off." It is almost never the connection and it is almost never the camera. It is usually the room, and the room is the last party in the building with any evidence of that.

Two habits fix the asymmetry, both free. First, believe the far end immediately — they have information you physically do not have, and the cost of checking is fifteen seconds of counting open audio paths. Second, ask early rather than waiting to be told: "how are we sounding?" in the first minute, addressed to a named person, gets you a real answer while the meeting is still young enough to fix. The counting procedure itself is on the room echo page.

The same asymmetry shows up without any echo at all. A room that answers a question nobody remote could hear looks, from the far end, like a room having a private meeting with an audience. That's why "the chair repeats the room" appears on half the pages here — it's not politeness, it's the only way the far end can tell that a thing was said.

Mute is not off, and it never was

The most expensive misunderstanding in hybrid meetings is that muting a laptop makes it safe. It doesn't. Mute closes the microphone and leaves the speaker playing the meeting into the room, where the anchor's microphone collects it and sends it back out — half a loop, which sounds identical to a whole one.

Zoom's own echo guidance makes this point in its own way, telling people in a conference room with multiple devices to disconnect computer audio from the other devices rather than simply silencing them. The word to use in your invite line is audio off, never mute — and it's worth being pedantic about, because the two words feel synonymous to everyone who hasn't watched a room echo for twenty minutes while five people insist they're muted. They were. That was the problem.

The anchor should be the most boring machine in the room

A good audio anchor does nothing all meeting. It sits mid-table, lid open, untouched, running one application. This is why the organizer's laptop — which shares slides, opens documents, alt-tabs, takes notes and gets closed when its owner stands up to point at the screen — is the worst candidate in the room, despite being the one chosen roughly every time.

Nominate someone who is attending rather than running the meeting, and nominate the same machine every week. A slightly worse laptop that is always the anchor beats a slightly better one that sometimes is. The reasoning in full lives on the Join Map.

An empty room proves nothing

Audio changes get tested in empty rooms, at eleven in the morning, by one person saying "test, test" into a laptop. Then they get deployed into a room containing six people, a keyboard, a ventilation system and a corridor, and behave completely differently.

Nothing about noise suppression can be evaluated in a quiet room, for the excellent reason that there is nothing there to suppress. If you change a suppression level, change one thing, live with it through a week of real meetings, and ask the remote participants directly — they're the only people in the arrangement who can actually hear the result. The same applies to any trial of cleanup software: run it in Tuesday's actual meeting, not in a demo.

Write the room's settings on a card, in the room

The most common audio "regression" isn't a regression. It's a person — a helpful attendee who changed a controller setting mid-meeting to fix something, and then left. Nobody afterwards knows what the room is supposed to say.

The fix costs an index card. Date it, put it inside the credenza or taped under the table, and record only the things that route a complaint to the right owner: which platform the room runs, whether it's Windows or Android if it's a Teams Room, the OEM and model, whether a DSP is present, and the suppression level you chose on purpose. A few facts and a date will settle arguments for years. A wiki page will not, because nobody standing in a room at 09:01 is reading a wiki.

"One audio path" counts rooms, not people

A recurring confusion worth flattening: the anchor rule is about shared air, not headcount. Eight people in eight different homes are eight rooms and eight entirely legitimate audio paths, which is why fully-remote meetings rarely echo. Two people in one room with two live laptops are one room with two paths, and will echo.

The corollary surprises people: when two offices join the same meeting, the correct answer is two anchors — one per room — not one anchor and a silenced site. Muting a whole office to "be safe" is a fix that costs you half your participants.