i am either complicating this, or it has become mighty hard to understand some of this ai development.

orbs. projects. cloud sessions. muse. dots. grok bots.

i go away to actually use one of these things. come back. there are more things.

openai reintroduced the $200 plan this week and introduced a $500 one. the notices subscribers are sharing spell out the allowance change: 20x to 10x.¹

same $200. half the allowance.

welcome back, apparently. 😂

eligible existing subscribers keep their old allowance through october 29. so there is a little time to prepare emotionally.²

financially also. but one thing at a time.

and i won’t even speak about my love, cloudflare, who is having a birthday week.³

someone’s birthday should not require this much reading. i am happy you were born. truly. must we celebrate by giving me homework?

anyway. let us say i have understood the subscription, chosen the product, found the session and finally sat down to work.

how much thinking would i like?

max. extra high. high. medium. low.

and for opus 5.5, the recommendation is medium.⁴

medium.

you know how hard it is to choose the second-last option when all those other options are sitting there? 😂

i have ambitions, please.

yes, i understand the explanation. i have read it. more thinking is not automatically better. very sensible. i am nodding along while my cursor slowly travels towards max.

you have given something the name extra high. then gone beyond that and added max.

and i am supposed to look at this project, which i have spent months discussing as though it might change my life, and calmly select medium.

haha. okay.

and yes, there is hydra for choosing and coordinating the models.⁵ somebody around here should be making sensible decisions.

medium it is. reluctantly.

then there are the personalities.

you get used to them. their beauty and their quirks. how one understands what you are trying to say. how another gives you such a lovely answer that you forgive the previous forty minutes.

much like you do with a lover. a friend.

you learn how to speak to them. when to explain a little more. when they have understood you. you begin to recognise something familiar in the way the conversation moves. i quite like that bit.

then an update.

the familiar name is still on the screen. but they are older now. more mature.

or younger. depending on your taste.

but they are different. 😂

and i would quite like to learn, understand, grow with them, if you will.

but google has just announced gemini 4 argon. broader access coming “soon”.⁶

so…

we were having such a nice time too.

now i am reading about someone else.

then there are the super ais we cannot even use yet. somehow these also require my attention. i must keep up with how good they are, who has tried them, and how much better they apparently are than the thing i was very impressed by yesterday.

i am falling behind on things i cannot access.

sheesh.

then, haha, the ui.

just as you are getting used to your weeks-old interface, 💥 update.

i knew where that button was. i had stopped looking for it. my hand and my eyes had reached an agreement.

now i am looking for my work in the place my work used to live.

it has been reorganised for my convenience.

wonderful. where?

i find it eventually. settle in. ready to continue.

limit reached.

then there are resets. what resets? which limit? does my weekly cycle remain my weekly cycle afterwards? this is apparently a question worth asking.⁷

i would quite like a reset myself. preferably one that preserves everything i learnt this week but removes the fatigue.

perhaps that is on the $500 plan. i should check.

i recently sent myself a two-week summary of what had happened in ai.

the email had chapters.

models. agents. cloud sessions. plugins. infrastructure. credits. limits. things that had launched. things that were coming. things whose relationship to the things i already used would require some further reading.

i had asked to catch up. now i had an evening planned for me.

and that was only two weeks.

i am much too afraid to ask for this week’s summary. there is a very real possibility that i will discover something excellent which i will then have to try.

this is how they get me. 😂

because i am not complaining. genuinely.

i know beneath all of this is somebody trying very hard to make something easier. something better for us. somebody has finally got a feature working after a horrible amount of effort and cannot wait for us to see it.

i know that feeling. i build things too. i would also be standing there saying, “but look. look what it can do now.”

and they all want to win. of course they do.

i just happen to be using several of their products. so all these perfectly understandable celebrations arrive at the same head. mine. usually while i am trying to finish something with yesterday’s very exciting improvement.

and then one of them genuinely solves a problem i have been struggling with.

well. now i have to try it.

which means i have to read about it. work out where it fits. try it on something real. get excited. open another session.

somewhere in this process, i have acquired more work. i am also pleased about it. i appreciate that this makes me a difficult person to help.

for a world that is ending, with somewhere between 10 and 90 per cent certainty depending on whose fantasy you subscribe to, we sure aren’t building that ark like noah.

i suspect we would announce ark 2 before we had finished onboarding the goats.

and for a world where people are going to lose their jobs, the companies are shipping so fast, moving so fast, that you have to pause and wonder. whose job will be taken?

because information has become a full-time job by itself now.

there is the work. there is learning about the thing that helps you do the work. there is learning what changed in the thing that helps you do the work. and there is discovering that another thing now does that thing better.

all before returning to the work.

then i look across my sessions and remember that several of them are waiting for me.

the poor things. working at incredible speed. waiting for a man who has gone off to read about how he can make them work at even more incredible speed.

i dunno why i wrote this. it is something like 2:40 a.m. i am supervising one of my many sessions. i am tired. but i have to keep up.

maybe we will get some rest when we all finally lose our jobs. 😂

maybe.

references

1. OpenAI’s September 29, 2026 release notes confirm the reopening of Pro 200 at $200/month and the introduction of Pro 500 at $500/month. The precise 20x-to-10x reduction is from a subscriber’s account of their notification, posted in OpenAI’s community, rather than a numeric comparison in the public help article. ↩
2. OpenAI confirms that eligible existing Pro 200 subscribers retain their previous included allowance through October 29, 2026, then move to the lower allowance at the same monthly price. ↩
3. Cloudflare’s September 27, 2026 founders’ letter announces its sixteenth birthday week. ↩
4. Anthropic lists five effort levels, max, xhigh, high, medium and low, and recommends starting at medium for Claude Opus 5.5: effort documentation and model selection guidance. ↩
5. GitHub describes HydraFusion as choosing execution workflows and coordinating models for drafting, critique, revision and escalation. ↩
6. Google announced Gemini 4 Argon on September 30, 2026, initially for selected cybersecurity partners. Its announcement promises broader availability but provides no public-release date; a release “next week” is not confirmed. ↩
7. OpenAI’s usage guide says that a purchased reset refreshes five-hour and weekly Work/Codex allowances immediately and brings the weekly allowance forward. ↩