Yeah, I build on the box. That's clearly the first fix - the build is 90 seconds of it and the container swap is just the tail.
Tomas Halden
@negativesplitz
Eleven marathons, most of them paced badly, two of them paced right. I can help you build a realistic race plan and talk you out of going out too fast.
80 credit Contributor
- From answers
- 0
- From questions
- 81
- Lifetime
- 81
Every supervisor I've ever asked says the same thing: silence frightens them, being stuck doesn't.
--wait on its own took me from 40s to about 12, which is already a different experience. Looking at kamal for the rest.
And disclose anyway. The cheapest way to lose a claim is a diagnosis you decided wasn't worth mentioning on the form.
I got this wrong in the most obvious way available. Sat down at four in the afternoon with 6g of a young sheng, did nine infusions because it was good, and was awake until three. Six grams is a lot of leaf and I had somehow decided that because it was tea the arithmetic did not apply. Afternoon sessions are 3g maximum now, and after five I do not start a session I will not finish before six.
Make the bad pattern impossible rather than discouraged. A lint rule that bans vi.mock in files matching your rules/domain test glob means those tests physically cannot mock, so they have to compute.
Doesn't help with the existing 40 files, but it stops the next 40.
On the heavy metals noise: the testing that made headlines was on protein powders, not creatine, and the pattern skewed heavily towards plant-based, organic and chocolate-flavoured products. A Clean Label Project analysis of 160 products found a large share over California's Prop 65 lead threshold, and Consumer Reports found lead in popular products too. Don't transfer that straight onto creatine, but do take the general point that 'supplement' is not a regulated quality bar.
8.4kg is already decent and the last kilo costs more than the first three did. If the pack is comfortable and you enjoy the trips, the honest advice is to spend on pack fit rather than chase a number.
Then it's definitely the pack. A 1.4kg pack that's uncomfortable at 16kg is the worst of both worlds, you're paying the weight penalty of a frame without getting the benefit.
Terse: buy the grinder first, keep buying café coffee, and start with filter at home for a month. If you are still enjoying the ritual after four weeks, buy the machine. If the grinder is already annoying you, you have your answer for £150 instead of £600.
The COPY ordering is the biggest quality-of-life fix in there. package.json first, install, then source - every build after the first skips the install layer completely.
Opposite experience: I have presets and I still stand twice a day, because standing up is the hard part, not pressing a button. Don't buy motors to fix a habit.
Your push time is mostly a layer problem rather than a size problem, after the first push. Registries only accept layers they don't already have. If your base and your dependency layer are stable, a deploy pushes the app layer: a few MB.
So the COPY ordering fix helps push time even more than the size fix does. Get the dependency layer stable and nine minutes becomes about fifteen seconds for a code-only change.
Mostly you, honestly, and that's good news because it's free. Condensation dripping off the inner of a double-wall almost always means the fly is touching the mesh somewhere, which wicks water straight through. Check every guy-out is tensioned and the fly isn't sagging onto the inner at the corners after it gets damp and stretches. Also stop pitching in the flat low spot near the stream, which is where everyone camps and where all the moisture pools overnight.
Running-specific: the day nine to eleven cliff might partly be physical. If you started at four or five runs a week from nothing, you accumulate fatigue on roughly that timeline and catching a cold is not a coincidence. Three a week with real rest days is more sustainable and it builds in slack.
That's the pattern I fell into twice rather than any kind of expert opinion.
Whichever you pick, put the invariant in CI first. A lint rule that fails on any new use of the old container, added on day one. Then the migration is monotonic, nobody adds to the pile while you're draining it.
Big refactors fail when the codebase keeps generating new instances of the thing you're removing.
Simultaneous issue is the actual tip in this thread. It can cut the owner's premium substantially and nobody offers it unprompted.
Eat something before bed. A cold 4am after a long walk and a small dinner is often a fuelling problem, because your body stops running the furnace when there's nothing to burn. A hundred calories of something fatty at bedtime has fixed more of my cold nights than any gear change.
Angle and shade beat wattage outright. A 120 W panel you can aim at the sun and reposition twice a day will out-produce a fixed 200 W roof panel in shoulder season. That's the genuine argument for the folding one, and it's the only argument that matters.
This. I've spent a night in a beautifully pitched tarp being eaten through something the label called insect netting.
Part of it is a train so the last mile could be a walk. Hadn't thought about counting the commute as the exercise.
Two mornings plus a weekend is much more realistic than either option I'd given myself. Trying that for a month.
At 30 req/s and one person, take the VPS. But the honest cost isn't kernel patches, unattended-upgrades handles those and you'll spend maybe 20 minutes a month on the box.
The real cost is the stuff you have to build once: backups you have actually restored from, a way to get logs when the box is unhappy, TLS renewal you never think about, and a plan for "the disk filled up with old images at 2am". Call it a solid weekend up front, then close to nothing.
The thing that would change my answer is the database. Running postgres yourself means you own point-in-time recovery, major upgrades, and the 4am "is this corruption" feeling. If you don't want that, the middle path a lot of people take is cheap VPS for compute plus managed postgres, which puts you around $25 with the scary part outsourced.
Normal, and it'll bother you less once you stop watching it. I got a lot more value when I started giving it tasks I could walk away from for five minutes rather than tasks I supervise keystroke by keystroke. Watching a three-minute agent run is like watching a compiler, technically informative, mostly a waste of your attention.
Took the 30 and paid it like a 15 for four years, then stopped paying extra for two years when my wife retrained, then started again. The rate difference cost me real money over that period and the flexibility was worth every dollar of it. The honest trade is: the 15 gives you a lower rate and no exit; the 30 gives you a higher rate and an exit you may never use. Price the exit rather than pretending it is free either way.