Ask

Is it normal for a first boot to sit on a black screen for 40 seconds

Completely normal. The first boot with a new DDR5 kit runs memory training, which is the board testing timings until it finds a set that works, and 30 to 90 seconds is an ordinary result. Two minutes isn't unheard of on four sticks.

The critical part is not interrupting it. If you cut power halfway through the training, the board can save a broken partial result and the next boot is worse. Set a timer for three minutes and go make tea.

178 · in/pc-builds ·

New build only shows 16GB of my 32GB kit after every cold boot

Separate from the training thing: DDR5 takes more force to seat than people expect. Open the case flat, push one end until the clip clicks, then the other, then press the middle of the stick until you feel it stop moving. If only one clip is latched the module can look installed and still lose contact when the board warms up and flexes.

163 · in/pc-builds ·

Textile jacket smells like a gym bag, can it go in the machine without killing the waterproofing

Armour out, all zips and velcro closed, technical wash rather than detergent, gentle cycle, no fabric softener, no bleach, low or no spin if the machine lets you. Detergent leaves a residue that attracts water and that is what kills the outer finish, not the washing. Then dry it and reproof the outer shell with a wash-in or spray-on product and it will bead again.

162 · in/motorcycle-gear ·

Ring or watch if the only thing I actually look at is sleep?

Wore a ring and a watch simultaneously for about four months, which is as ridiculous as it sounds and did produce an answer.

Ring: I forget it is there, which is the entire point for sleep. Battery in days rather than hours so it charges while I shower. Sleep and overnight metrics were close enough to the watch that I would not choose between them on accuracy.

Watch: better at everything that happens while awake, and it is a watch, so it tells the time and buzzes.

For your use case, which is sleep plus lifting, the ring is the obvious answer and the watch is the thing you already do not want to wear.

213 · in/wearables ·

Session cookie works on localhost but vanishes the moment I deploy

Fastest way to confirm the diagnosis before you change any code: open the app, then look in devtools under Application, Cookies, and select the API's origin rather than the page's origin. If the cookie is sitting there quite happily but not being attached, it's a SameSite or credentials problem. If it isn't there at all, it's a Secure or Domain problem. Two very different afternoons.

341 · in/fullstack ·

CORS preflight passes in curl but the browser still blocks my PATCH

Worth knowing that the CORS headers have to be on the real response as well as the preflight. A lot of setups add them to OPTIONS only, then the PATCH itself succeeds server-side and the browser throws the result away, which looks identical from the front end and is very confusing when you check the database and the row has changed.

205 · in/fullstack ·

Book a studio day for drums or keep programming them at home?

Disagreeing with the pay for a day consensus. Modern sampled drums with a real drummer performing them on pads or programmed carefully with velocity and timing variation are genuinely fine, and half the records people cite as sounding live have heavy sample reinforcement anyway. If your budget is one thing, I would put it into someone mixing the record properly. A great mix of programmed drums beats an average mix of real drums recorded in a rush.

127 · in/home-recording ·

AI chat free tiers cut me off after a handful of messages, rolling window or a fixed daily clock?

Worth saying before anyone measures too carefully: these caps change without announcement, sometimes weekly, and often differ by region, by whether you're signed in, and by which model the request lands on.

I kept a note of what I'd measured for one provider and it was wrong within a month, not slightly wrong, half. So measure to learn the shape of the window, don't build a routine around the exact number, and definitely don't build anything automated on top of an undocumented free tier.

152 · in/free-tier-limits ·