October 2026
Q4 is thirteen weeks, and one outcome decides it: the one school left on Sabu finishes a full term and comes out of it ready for paying customers in January.
Sabu
The Anglo-Arabic school was the pilot, and that ended in the last week of September. So the pilot is now the Quran memorisation school, the only one on Sabu.
Everything bends around its calendar. Grades and report cards have to work before exams start, so anything that feeds a report comes first and UI polish fits around it.
The exams and reports get built either way. If no school is running a term through them by December, they still get built, because January needs them.
October is for locking scope. Every remaining feature gets marked “needed this term” or “after pilot”, the exam and report-card dates get confirmed, and the gaps the school feels daily — attendance, fees, class lists — get closed. There is a twenty-minute check-in with the school contact each week, which is the thing I am most likely to let slide and the thing most likely to save me.
November is for hardening what they actually use: grade entry and report cards against their real data, pilot bugs fixed in the week they are reported, backups verified before the reporting period starts.
December is the end of term. New features freeze two weeks before reports go out. Then a retrospective with the school — what worked, what is missing, and whether they would pay.
Teaching
Web I and II, Database I and II, on Fridays and Saturdays. Every remaining week gets mapped to a topic now, exam papers are set by mid-November so December isn’t crowded, and marks go back within a week of each assessment.
Kill list
Friday and Saturday mornings, seven to half past eight — twenty-six sessions this quarter, one project at a time. The offline sync engine for Sabu stays first until it works: local store and change log in October, sync and conflict handling in November, then testing under pilot-school conditions, which means a poor connection and shared devices. Nothing else starts until sync ships.
The shape of the week
Sunday is the big Sabu day, and ends with a twenty-minute review at four. Monday to Thursday belong to UTG, nine to four, and the rule I am testing is that the work stays inside those hours — whatever is left goes on the next day’s list, not into the evening. Exercise is Sunday to Thursday at half five, sixty-five sessions by the end of December. Evening blocks start after Isha.
There is no full day off in this. Friday and Saturday evenings are the rest, and one Sunday afternoon a month.
Turning over
The guardrail I expect to be tested first: anything new that arrives — a client, the WISP, another project — waits for the Sunday review instead of walking into the week. And if Sabu falls behind, UI polish gets cut before anything the school needs for reports.
The school before this one was still on it the week before it ended. So this is the opening position, not a forecast.
Who can judge the code
Proton’s CTO wrote an engineering manifesto this month, published by 404 Media on 29 October: an engineer’s job is no longer to write code but to help the model write better code faster, every engineer is a team leader whose direct reports are the models and agents, and most human code review will eventually end. The direction I don’t argue with. The guardrails I do — a model reviewing a model’s output shares its blind spots, and at a company whose product is security that is the part users never agreed to.
What stayed with me is the question underneath it. Today’s seniors can tell good code from plausible code because they spent years writing it badly and then fixing it. If firms stop hiring juniors, and the juniors who are hired never hand-write anything, who is left in ten years to catch what the reviewing model misses?
I am standing at the start of that pipeline. I don’t know yet whether the answer is hand-written work before AI tools are allowed, or marking students on critique — find the planted bug, explain the flaw, justify the fix — or both. But “explain this line by line” is starting to look like the assessment rather than the extra.
I wrote the argument out properly the next day, in The manifesto needed the engineer it mocked.