Anneke Roos

Allowing insecure content for the catalog domain in Edge did it. Download popup opens, file comes down. Amusingly the file itself still arrives over https once the permission is set, the block seems to be about how the popup is constructed rather than the transfer itself.

Fifteen months is not that long, be slightly kinder to yourself. The N4 to N3 gap is roughly double the vocabulary, so part of the flat feeling is just that the next step is bigger than the last one.

if you do want it fixed cheaply, one scheduled job at eight on monday morning that opens a connection and runs select 1. costs nothing, runs once a week, and the first human of the week gets a warm path.

Also check for blank cells inside the source range, not just at the bottom of it. And make sure the pivot source is a table or a dynamic range — if the range says A1:F3000 and you now have 4,000 rows, the pivot silently ignores the rest and you will get stranger problems later.

Rebuilding does not fix text dates, it just makes you do the work twice. Fix the column first.

Bitrate, and specifically you are probably on constant bitrate with a high target and an even higher ceiling. For a ten minute 1080p25 deliverable that is mostly people talking, two pass VBR at a target around 8 to 10 Mbps with a maximum of 12 to 16 lands you near 700 MB and looks identical on any normal monitor. 1.8 GB over ten minutes is roughly 24 Mbps, which is broadcast territory for content that does not need it. Also check your audio is not sitting at uncompressed PCM — that alone can be 100 MB on a ten minute file.

Two pass VBR costs you export time and nothing else. Use it for anything a client has to download.

I would go further than the others and say that two of your three options are not building in public at all, they are content about building. The practice, as originally meant, was that the decisions and the numbers are visible so people can learn from them and hold you to them. If you strip out the learnable part you have a marketing channel with a nicer name, which is fine, but then judge it as marketing and it usually loses to just being useful in a forum where your users already are.

Just wash it twice, problem solved.

We tried an app-based service first because it was cheaper per month, then moved to an in-person therapist. The app was fine for structured exercises and useless for the actual argument, because the sessions were short and we got a different feel every time we had to re-explain our history. The in-person work cost more per hour and needed fewer hours. If money is the constraint I would do fewer sessions with one good person rather than more sessions with a rota.

It happens every time someone copies and pastes formatted rows. Delete them all and reapply once to the used range. Check the used range itself while you are there — Ctrl+End sometimes lands on row 900,000 because of old formatting, and everything is slower for it.

module scope boolean, flipped at the end of the first handler run. log it on every request as cold true or false along with the age of the instance. ten lines, works on every platform, and it survives you changing providers.

Not normal. 8,000 rows should feel instant, so something in there is either volatile or scanning whole columns. The order I check:

• Formulas referencing A:A instead of A2:A8000. Fifteen columns of SUMIFS over full columns means a million row scan per formula, per recalculation.
• Volatile functions: OFFSET, INDIRECT, TODAY, NOW, RAND. Any one of them forces the whole workbook to recalculate on every keystroke.
• Conditional formatting applied to entire columns, or rules that multiplied when people copied rows. Open the rules manager — I have seen 4,000 rules in a sheet that needed three.
• Array formulas or SUMPRODUCT over full columns, same problem as the first point.
  Fix those and it drops under a second.

Power numbers, since that is what decides it for outbuildings on a long cable run: a Pi 4 sitting idle pulls somewhere around 3W and spikes higher when it does anything, while an ESP32 doing a read-and-sleep cycle averages in the milliwatts. If you ever want this on a small solar panel and a battery, that gap is the whole decision. On mains it does not matter much, but the Pi still needs a clean shutdown and the microcontroller does not care.

Exactly. Anything with a moving part attached to a living animal should be the boring device you never log into.

Make the toile close with the opening at the front, even if the finished garment opens at the back. You can then adjust the back on yourself by pinching and pinning at the side seams and shoulders where your hands actually reach, then transfer those changes to the paper. Fighting a back opening on yourself is how people conclude that solo fitting is impossible.

Leak sensors keep coming up and I own zero. Ordering three.

Anything tied to a clock rather than to a person eventually becomes rude. That is going straight into my notes.

The using locale option was the whole thing. It parsed the trailing minus correctly with no formulas at all.

Be aware of the hardware ceiling before you pick: on a Pi 3 the ethernet port hangs off the same USB 2 bus as the storage, so realistic throughput once you are also reading from a USB drive is well under 100Mbit. That rules out file server, media transcoding and anything backup-shaped where you care about the speed. It does not rule out anything that moves small amounts of data, which is most of the genuinely useful stuff.

Labels at both ends of every planting group and a photo on your phone in May. That is the entire system.

lazy require the heavy things, keep module scope for pure definitions, and stop reading config at import. that is most of it.

One more argument for consolidating: INDIRECT is volatile, so 40 sheets of SUMIFS built on it will recalculate on every keystroke. On a workbook this size you will feel it, and it gets worse every month you add.

Confirm it is actually hum before you chase the room. Look at the spectrum: mains hum is a hard line at 50 or 60 with harmonics stacked above it. If it is broad and low instead, you have a different problem — clothing rub, a failing connector, or the lav power supply itself.

For the text-number version of this specifically, the fastest one-off fix is: select the column, Data, Text to Columns, leave every default alone, press Finish. It converts them in place in about two seconds. Ugly, and it does not survive the next import, but it unblocks you while you sort the import out properly.