Aussie iOS dev living, working and playing in beautiful Whistler B.C. Tweets are my own.
Whistler, British Columbia
Joined September 2008
- Tweets3.4K
- Following247
- Followers1.3K
- Likes6.2K
Ian Keen retweeted
>sam altman: "we tried this because there were rumor on the internet"
>rumor: math professor close to solving navier-stokes
>the professor: using codex for a year
>openai: has all his logs
>checked every codex session tagged to N-S
>found the most promising one
>spun up 10,000 agents to finish the proof where he couldn't
>they got caught
>told the guy if you say anything you'll destroy your own career
>he ruins his career over this
>openai: "we did not see his work"
>also openai: "we cannot rule out that de-identified data from their usage helped improve our models"
they stole it lmao
It’s not that hard y’all.. you can one shot anything with simple prompts
“build me a game called ‘slop town’ reference the code in ‘~/existing code/existing game’. Make it 20x better”
Coding is solved.
Ian Keen retweeted
ever since AI got smart enough to do shit autonomously i swear it has only made me dumber and lazier
i cannot even get shit done anymore
all i do is sit there staring at my screen watching AI do everything
and is this really the same shit that was supposed to make humanity immortal and push us forward???
like bro what the fuck
a few months ago i was moving between websites, opening 30 tabs, scraping for information and actually finding shit myself
now i just type something into a box and wait
there is no fun in this man
i do not want to watch a machine use the internet for me
i want to do shit with my own two fucking hands again
i want to search, get lost, figure things out and actually feel something while i am online
fuck being 10x more productive if i am not even the one doing anything
i do not want it anymore man
Type out what to add plus prose.. burn time and token for it to add what you already typed.. well.. I’m convinced, we’re living in the future
Ian Keen retweeted
I'm seeing team leads under immense pressure.
Why? Because AI decoupled effort from output.
Developers can create giant PRs in a few minutes. But that code may take hours to vet, test, fix, and review.
In 2024, bad developers were easy to spot. They produced little code, or struggled to write code that even ran.
In 2026, bad developers can use AI to quickly generate tons of code they don't understand. It seems okay at a glance, but contains subtle problems and fundamental issues that are hard to spot.
"The energy needed to refute nonsense is an order of magnitude bigger than the energy needed to produce it."
-Brandolini's law (aka the bullshit asymmetry principle)
Ian Keen retweeted
just one more harness bro. i promise just one more harness and we'll go from slop cannon to software factory bro. it's just a more powerful harness bro. please just one more.
Ian Keen retweeted
AI has completely changed what I loved about software engineering.
Earlier, you’d sit with a problem, understand it deeply, write the code, break it, debug it, and eventually own the solution.
Now so much of engineering feels like staring at a screen waiting for an agent to finish.
And the worst part isn’t even that AI writes code. It’s that we’re slowly losing ownership of the codebase.
AI generates thousands of lines we barely read. The code becomes a black box. Then something breaks, and suddenly you’re expected to debug a system you never really understood in the first place.
At the same time, constantly delegating thinking to agents feels like it’s destroying our attention span and our ability to patiently read, reason about, and understand code.
Maybe we’re shipping faster.
But sometimes it genuinely feels like we’re becoming worse engineers while our codebases become harder to understand.
Ian Keen retweeted
Read the code. Every day there are examples. You cannot have enough guidance to hit every edge case. I do no believe you.
Ian Keen retweeted
my take is that ai has made it almost impossible to tell how much care someone actually put into a pr
i don't necessarily care if ai wrote all of the code. i care whether you understand it, tested it, refined it, and stand behind what you're asking me to review and ship.
when the code, description, and even responses to review comments are generated, how do i know you did anything beyond generate and move on?
if you're asking someone else to carefully review your work, you should have carefully reviewed it first.
Ian Keen retweeted
Engineers who really care about the quality of their teams work are getting burnt out. Overwhelmed by PRs, difficulty in explaining why implementation choice maters, noticing that its a much easier policy to not care and slop it up. Wonder if it's a pattern..
Ian Keen retweeted
software engineer's greatest power was how collaborative it WAS. but with 200 lines PR descirptions, bloated PRs, design docs which are made by AI for AI and little to no discussions apart from YOLOing with codex or claude code, we are fundamentally at a place where all of us are just talking to our agents and not each other.
and the strange part is that with more docs than we have ever had, longer PR descriptions, cleaner commit messages, green dashboards, faster cycle times, every proxy we ever used for a team thinking well together is going up while the thing underneath it is quietly going somewhere else, and the same thing happened to review, with diffs too large to hold in your head, with summaries generated on one side and summaries generated on the other, with the reviewer prompting instead of reading, we are in this slightly absurd position where two people sit on either end of a conversation that neither of them is having.
we are fundamentally somewhere new at 2am when something breaks and there are three stupid reconstructions of the same subsystem and no way to tell which one is remembered and which one is generated(are we being deadass serious?).
Juniors(well there aren't any juniors) are inheriting all of this as the default state of the world, with no senior tearing their PR apart in public, no two day sting(115 comments on a 50 lines PR), no argument they lose and then think about on the train home, just re prompting until it goes green, we are somewhere i do not have a name for yet, where the loop teaches you how to ask again rather than how to think again.
i do not know what any of this settles into, but with all of us fluent in our agents, less fluent in each other, and mostly fine with it, that is roughly where we are.
