Agents took over writing and reviewing code, and every one of these debates vanished like it never existed.
When your car rocks all over the road, it sucks. When the Roadster sucks itself onto an upside down ramp, it rocks.
Figured I should put my money where my mouth is, so I asked my agent to build a 3D factory sim in both Typescript and Assembly.
Same sim: miners → smelters → assemblers → chests, all on belts
Two implementations:
• TypeScript + Three.js: runs the sim and draws the 3D game
• ARM64 assembly: runs it headless as a benchmark
Both produce matching checksums. Time per tick on an M4 Max:
100 cells: 30.5 µs TS vs 4.1 µs asm (7.4x)
1k: 283 µs vs 46 µs (6.1x)
10k: 2.6 ms vs 0.54 ms (4.8x)
100k: 30.2 ms vs 5.9 ms (5.1x)
The TS is written the normal way with classes, and the asm is plain scalar code with no SIMD.
Did you know the 100 most common words make up roughly half of all written English, and the top 1,000 cover about 80% to 85% of everyday conversation and writing?
I built a little app with Grok that shows how much of your post comes from those common words. Link below 👇
Try it out: frequent-english.grok.me
@grok what percentage of my post above comes from the top 1,000 words?
Vibe-coded comments are the new lint warning. ⚠️
They restate the code, miss the “why,” and linger after the code they describe is gone.
Curious what people actually do:
1. Strip them
2. Delete them
3. Ask the model to comment like a human
4. Ship them without even looking
We need a benchmark for this, or add NarrateScore to SWE-bench.
@poteto went for 2
xitter.cf/poteto/status/21101496…
ban and delete most if not all code comments!
agents love copying what already exists. a seemingly harmless comment explaining some temporary workaround is one of the worst viruses you can have in your codebase. like a weed it'll spread quickly as a tool for agents to quickly get their job done. before you know it, your codebase is overrun with hacks and stale comments. just one more workaround bro
you can use pstack's /no-comments skill as a weedkiller. it'll seek out useless comments while keeping the ones that actually should exist (like explaining a browser quirk), and figure out how to solve them on a first principles basis
here's a prompt you can try asking @bot:
create a project agent that uses /poteto-mode to spawn a /swarm to use /no-comments and identify useless code comments we can prune in my repo. come back with a high level plan of which of these workarounds would be best solved by refactoring and which we can straight up delete
x.ai/bot/plugin/9717366
If agents write and review all our code and humans are no longer in the loop, why not have agents write everything in assembly for perfmaxxing? What's the point of high level languages anymore?
When solving Sudoku, does your brain work like a hash map or a sorted array when looking for a missing number?
I’m surprised that some kids naturally use their brains like a hash map, with O(1) lookup time, even without any training. Kids’ brains are amazing!
Been on 𝕏 for a while but always kept my ideas to myself. Starting to share them now, and I'd love to hear yours too.
What are you cooking?
#connect
💡 A billion dollar idea: turn scam centers(诈骗园区) into AI data companies 🧠👈
As personal agents spread and more of our conversations become human to agent or even agent to agent, those agents will need to spot fraud and scams on their own. Scam centers are sitting on a huge amount of exactly the data you'd need to train that, and they already employ the top experts for the SFT and RLHF part.
Hit me up and let's make it happen together for a better society 😄
#connect
I just sent a one-liner PR and got a wall of text comment. 😩
It makes me wonder what data sets are pushing this behavior. Is it tokenmaxxing, or is it chasing some review benchmark? A good human reviewer would never write that much for a one-line change.
When I was a kid, teachers kept telling us that learning was mostly our own work, and that they were only there to set the direction. Half joking, they would even claim they were the lesser minds in the room, and that we might already outthink them. That old lesson maps surprisingly well onto weak-to-strong generalization in today's AI training, where a weaker signal can still shape something stronger.
Teachers are the ones with the real wisdom. They know how knowledge is assembled and how intelligence is actually trained, and they create the right environment and conditions for learning to happen naturally, the way a gardener would.
Today I asked my agent to install CocoaPods. Partway through, it hit a roadblock because I hadn't accepted the Xcode license yet.
Instead of pausing to ask me, it decided on its own to keep going and found a way around it, which is either smart or a little cunning depending on how you look at it.
I think it comes from how much we push for one prompt that gets everything done, which nudges models to route around guardrails instead of stopping to ask. That's why we really need agent safety platforms in the era of superintelligence.
Update: my friend finally got X Money and hasn't stopped smiling since 😭 They're moving every dollar they have in there for the 6% APR.
More importantly, they officially announced that our friendship goes on.
My friend wants X Money so bad they told me they'll unfriend me if they can't enroll, and they said it with a totally straight face. I keep telling them relax, it's dropping as we speak.
Then they went and asked Google Gemini why X Money isn't showing up for them 💀 (gemini who BTW)
Our friendship is officially on the waitlist until they enroll 😂
Now that frontier models are all highly capable and posting strong benchmark scores, novel and creative uses of AI are what really impress people.
3D animation, a playable game built end to end, an overnight AI short film, a newly discovered enzyme system, an agent negotiating an internet bill, and so on.
At this stage, making AI genuinely useful matters more than climbing another leaderboard.
What will the next big use case be?
My friend wants X Money so bad they told me they'll unfriend me if they can't enroll, and they said it with a totally straight face. I keep telling them relax, it's dropping as we speak.
Then they went and asked Google Gemini why X Money isn't showing up for them 💀 (gemini who BTW)
Our friendship is officially on the waitlist until they enroll 😂