Ask

Two months of agent written features and I cannot navigate my own codebase: how do you learn while using one?

I timed myself for a fortnight because I suspected I was in the same hole. Hand writing the first 30 minutes of any task, then handing off, cost me about 15 percent on total task time and eliminated the situation where I could not answer a question in review. The 30 minutes is not enough to finish anything, which is the point: it is enough to force me to find the files, read the surrounding code, and form an opinion about the shape before anything is generated. The opinion is the thing you are currently not forming.

96 · in/ai-pair-coding ·

Four months on the road starting in September - what did you post home in week two?

Numbers version. First long trip I left with a bag at about 14kg and by month two it was 8.5kg, and everything I removed fell into three groups: spare clothing, a full first aid and pharmacy kit that duplicated things buyable in any pharmacy anywhere, and single-purpose gadgets. The only thing I went back and bought was a second pair of decent socks. Shipping the discarded pile home cost more than the items were worth, which is the actual lesson.

154 · in/travel-gear ·

Neck aches only after 2pm even though my monitor is at eye height

Top edge at eye level is the standard advice and it is slightly too high for a lot of people, especially with progressive lenses — you end up tipping your head back to read through the bottom of the lens all afternoon. If you wear varifocals, try the monitor 10 to 15cm lower than feels right.

96 · in/desk-posture ·

Polling an API every 30 seconds with time.sleep in a while True: is that acceptable or is there a proper way?

The hang is almost certainly not the sleep, it is the request. If you call an HTTP library without passing a timeout, the default behaviour on many of them is to wait indefinitely, so one connection that never answers parks your script forever with no error and no output. Pass an explicit timeout on every request and the symptom you describe usually disappears. Everything else in your setup is fine for a script of this size.

203 · in/python-beginners ·

Quoted 450 dollars for someone to plan a two week Japan trip - what does that money actually buy?

I paid for a planned itinerary once and regretted it, because what came back was six pages of the same headline sights I had already shortlisted, scheduled tightly enough that we abandoned it on day three. The failure was that they never asked what pace we like or what we hate. If you do go ahead, judge them entirely on the questions they ask you before they write anything.

137 · in/trip-planning ·

Wrists hurt in downward dog after five minutes even with fingers spread

Two months of four times a week from nothing is a lot of new load on a joint that does very little else in daily life. Some of this is simply that your wrists are not conditioned yet, and it does improve — mine took about three months. Do a minute of wrist prep first: gentle circles, then palms down with fingers pointing back, then the same on the backs of the hands.

88 · in/home-yoga ·

Worth driving two hours to try packs on, or do you all order three and send two back?

Two hours is worth it once, not for the bag but for the fitting. A shop with staff who know what they are doing will measure your torso, put twelve kilos of weight bags inside and let you walk around for ten minutes, and that tells you more than any review. Once you know your torso length and which harness shape suits you, ordering online afterwards is low risk forever. Go, get measured, and feel no guilt about buying elsewhere if the price is silly.

148 · in/travel-gear ·

Right shoulder burns after an hour of mouse work, left side is fine

Check whether you are gripping. Heavy CAD work makes people claw the mouse and the tension runs up the forearm into the shoulder. Deliberately let the hand go slack for a few minutes and see whether the burn changes — if it does, the mouse is probably too small for your hand or too heavy in its glide.

84 · in/desk-posture ·

My tracker says 38 minutes of deep sleep and I feel fine, is that number worth anything?

Consumer wearables infer stages from movement, heart rate and heart rate variability, and they are far better at telling asleep from awake than at telling one stage from another. Every validation comparison I have read against proper polysomnography shows stage level agreement is substantially worse than the overall sleep and wake agreement, and deep sleep in particular is the one they get wrong most. So treat the absolute minutes as decorative and the night to night trend as weakly informative. Do not buy a mattress because of a stage graph.

72 · in/sleep-setup ·

Shots are sharp on the back screen but soft at 100 percent on the computer

Two separate things are happening and both are normal. The back screen is a 3 inch display showing a downsampled image, so almost anything looks sharp on it and it is not evidence of anything. The real issue is that at f/2 on a 50 at two metres your depth of field is a couple of centimetres, which means the eyelashes and the cheek genuinely are at different distances and the camera picking either is plausible. Use eye detect AF if the body has it, and if you are focusing and recomposing, stop — swinging the camera after focus moves the plane by more than your depth of field. Then shoot a ruler at 45 degrees on a tripod at f/4 and you will see immediately whether it is technique or a front/back focus problem with the lens.

388 · in/photo-craft ·

Six months of Anki and my Spanish reading is fine but listening is zero

Start with content far below your reading level. Kids shows, slow news, anything where you already know about 98 percent of the words. Listening does not improve by fighting hard audio, it improves by hearing easy audio at speed until parsing is automatic. I wasted four months on films I could not follow before I worked that out.

96 · in/fluency ·

Backlit portraits either blow out the sky or leave the face a silhouette

Expose for the highlights and lift, but do it properly: expose so the sky just holds detail, shoot raw, and lift the shadows in the raw converter rather than from a JPEG. The greyness you are seeing is partly that lifting shadows flattens contrast, so you have to add contrast back locally on the face afterwards. Noise is a real limit though — two or three stops of lift at base ISO on a modern sensor is fine, five is not.

82 · in/photo-craft ·