new rule, for three months: after the last tuition, the phone goes into the almirah until i leave the next day.

the scrolling was eating the evenings, and the evenings were eating everything else.

four pages of the quran a day instead. one focused thing at a time. we'll see how long it holds.

skipped tuitions for a chess tournament. one day, eight rounds. my first.

five wins, a draw, two losses. 29th of 212. five of my opponents had fide ratings.

i started playing properly in september. it's the first thing i've done in a long time where the board tells me exactly how wrong i was, every single move.

found out on thursday that the marathon started at 2:45 on friday morning. started three minutes late.

the shoes were too tight from the first loop. changed them at 33 km. it didn't help. someone rubbed my legs at 38, and the last two kilometres were my fastest of the night.

4:58. my first marathon under five hours, 32 minutes off my best. the legs weren't the problem. the feet and the stops were. that's fixable.

ran 35 kilometres today. the longest i've ever run in training.

the first twenty went the way long runs go. the last stretch was mostly negotiation. i got home weaker than i expected to be.

two weeks to a marathon. this was supposed to be the confidence run. it turned out to be the honest one: the distance is there. the margin isn't yet. taper starts now.

the last article went up today. keyboard to screen — one keystroke followed all the way from a switch closing under a finger to photons leaving the glass. the series is finished.

seven weeks ago i wrote in this same field-notes file that all eight were done on paper and i was scared to publish them. that entry is still below this one. i left it there on purpose.

what i got wrong back then: i thought the articles needed to be finished before they could go up. they weren't finished. i corrected a scheduler that linux replaced in 2023, a dram refresh interval, an l1 latency figure, and — in this last one — a claim that a keyboard reads all 104 keys on 104 wires. it doesn't. it scans a grid. that correction came after publishing, not before, and the article is better for it.

the thing i was afraid of — being wrong in public — happened, several times, and the cost was an afternoon of edits each time. the cost of waiting would have been the whole series sitting unread in a folder.

i don't think i know computer science now. i know where the map is. that was the whole point.

the ssc results came out yesterday. i'd built a site that ranks the chattogram board's results, and last night was the only night it mattered.

ran the scrape, checked a sample by hand, then published. by this morning students i've never met were looking up where they stood.

it's the least glamorous thing i've built and the most useful one to strangers. no login, no design award. just a number someone was waiting for.

the sylhet marathon is tomorrow. i'm not going.

a week ago, with the toe still sore from football in taper week, i wrote down — in advance, while calm — the conditions under which i wouldn't start. yesterday they came true.

every other time i've written a rule like that, i broke it the day it cost something. this is the first one that held. the trip went with it.

still annoyed. but the 100k is in december, and i'd like to reach it on two working feet.

the series is done on paper. all eight articles written, most of them sitting in the code quietly — not linked anywhere, not visible to anyone.

only the intro and article 1 are live. the rest exist but i'm not sure they're ready. or i'm not ready. or i don't want to release something and then have to live with it being wrong in public.

i don't know if staggered publishing is discipline or cowardice. maybe both. maybe the distinction doesn't matter if the articles actually get better while i wait.

the paper oscilloscope is what i'm calling it — warm aged-paper aesthetic, bangla-first. the first honest thing i've written about what computer science actually is, from someone still inside the confusion of learning it.

the tuitions moved to 7am this month. which means waking at 4:30, 4:45. which means fajr is no longer a prayer i choose to get up for — it's just there, at the edge of waking, unavoidable now.

i thought this would fix my namaj problem. for months my record was bad, more missed than kept. now i catch fajr almost every day. the getting-up is solved.

but i stand in it half-asleep, reciting on autopilot, my mind already on the 7am student. i catch the prayer and miss the prayer at the same time. presence turns out to be a separate problem from attendance. i solved the easy one and called it the hard one.

no clean lesson here. just noticing that showing up and being there are not the same thing, and i only fixed the first.

finished my first paid website this month. a real client, a real event. i poured more into it than the fee justified — made it as premium as i could, because i couldn't help it.

someone outside the project saw it and called it premium, amazing. an hour later the client sent a list of corrections. no thank-you first, just the list. it landed like a verdict on the whole thing.

it wasn't. the list was four small items, most of them fair. but for an hour i couldn't tell the difference between "here are some changes" and "you failed." that gap — between what the work is and what one blunt message makes it feel like — is the actual job. not the code.

the work was good and the work needed changes. both true at once. learning to hold both without the second erasing the first is the thing i'm actually practicing now.

2:32:40. six-oh-seven per kilometer. 68th overall out of 942. my first 25k back after the injury months, and somehow my most even race — the second half almost exactly the first.

it fell apart in one specific place. right glute gave out around km 18 to 20, and everything downstream started compensating. i finished on borrowed mechanics. the time held; the body filed a complaint.

the even splits felt like discipline while i ran them and like luck when i looked back. maybe they're the same thing at this distance — holding a pace you're not sure you can hold, and being wrong about the doubt.

woke up feeling like fever. body weak. it was a long-run day, but i decided to skip the morning and rest. tuitions all afternoon. the heat outside was worse than my head, somehow.

at 7pm i went out anyway. ten kilometers, slow as i could run. the weather was hell — humid, still warm. but i finished, no pain, just sweated through everything.

the lesson, maybe: my body lies to me about what it can do. the fever-feeling in the morning was real, but it wasn't a no-running signal. it was a be-careful signal. those are different.

today was a running day. but relatives stayed at our place, and a 5am dawn run would have raised questions. why is he out so early? where is he going? what is this?

the answer "i'm running a marathon next month" doesn't fit into the kind of conversation a relative wants to have at 6am tea. so i skipped.

this is what order-vs-truth looks like in practice. not big philosophical decisions about disclosure — just small ones. the run is moved to tomorrow. the relatives don't know. the order holds.

i listen to podcasts on my phone in the traffic between tuitions. usually background. occasionally a line catches.

today, ramjan bhai on a podcast with sadman sadik:

আমি তাদের ক্ষমা করে দিতে চাই, কারণ আমি চাই আল্লাহও আমাকে ক্ষমা করে দিক।

i want to forgive them because i want allah to forgive me.

played it back three times. it shifted something. forgiving the people who've wronged me has always felt like a transaction i wasn't getting anything back from. this reframes it: i forgive not for them, but for the kind of person i want to be received as.

simpler now. less argued with.