croemer

8 ms·

121.
by croemer 6mo ago
Commit history looks vibe coded doesn't it? Don't read too much into anything you see there. It's what Claude or Codex wrote after being asked to solve the challenge.
122.
by croemer 6mo ago
The person who won the challenge with this apparently misleading code seems to have absolutely no quantum computing background. He writes this as background about himself: > Technology leader with 10+ years in enterprise software, full-s
123.
by croemer 6mo ago
I've been getting lots of refusals by Codex with GPT 5.5 for "biosafety reasons" when asking for harmless things like code to analyze SARS-CoV-2 sequences for breakpoints. That's in no way useful for creating viruses wha
124.
by croemer 6mo ago
It refuses to write bioinformatics code that involves analysis of SARS-CoV-2. Even when it's totally obvious I'm not trying to do any bioengineering of any sorts. Totally harmless stuff I'm doing and I just get rejected.
125.
by croemer 6mo ago
Here's a demo video: https://youtu.be/m7mYzSZdNPE?t=21&is=oqvHiwmoyVujl_3Z
126.
by croemer 6mo ago
Ironic that the article seems to be at least partially LLM generated. Lots of negative parallel constructions (it's not x, but y) and "not merely". Also short sentence bursts and colons. Also, Pangram says 100% AI generated (
127.
by croemer 6mo ago
Reality is rarely an ideal monopoly or perfect competition. In practice, there's always a bit of monopoly power.
128.
by croemer 6mo ago
Would be good to show results in details, not just summary. A list of high and low scoring submissions, for example, to vibe check the method. Also, would be good to show trends over time rather than just a one-time pie chart showing breakd
129.
by croemer 6mo ago
It uses "not merely" 7 times! I wonder if an LLM would repeat a phrase that often. It could be the author starting to write like LLMs, instead.
130.
by croemer 6mo ago
That's not what Jevons paradox means though. He's just name dropping some concept. Jevons paradox would be if despite software becoming cheaper to produce the total spend on producing software would increase because the increase i
131.
by croemer 6mo ago
It's a figure of speech called metonymy. I agree Brussels is not very precise, a better word would be Berlaymont to refer to the EU commission specifically as there are a lot of institutions that could be meant by Brussels (Belgian fed
132.
by croemer 6mo ago
It does provide value in that I know I shouldn't read it. It's clearly LLM written after a few glances.
133.
by croemer 6mo ago
Why didn't you put the methodology in the post? Also, which devices were used to record? How do you know people went to sauna?
134.
by croemer 6mo ago
Use desktop view
135.
by croemer 6mo ago
What does "major update to codex" mean? New model? Or just new desktop app? The announcement is vague.
136.
by croemer 6mo ago
If it wasn't sloppiness they would have issued the directive on January 1.
137.
by croemer 6mo ago
Rules that are not enforced are bad as they create space for arbitrariness and corruption. It was a mistake by gov't, opposition & media that this wasn't spotted at the time the law was revised. The most surprising thing is th
138.
by croemer 6mo ago
They are a UK government unit: "The AI Security Institute is a research organisation within the Department of Science, Innovation and Technology." Unfortunately, they fit straight lines to graphs with y axis from 0 to 100% and x a
139.
by croemer 6mo ago
Yep, the LLMisms are there. "It's not X, it's Y", "Here's...": > The codebase doesn’t care how it got written. It cares whether it works, whether it can be maintained, and whether it helps the business
140.
by croemer 6mo ago
That's pretty much Switzerland. Indirect democracy for most things, but if enough people disagree with what the government does or they feel strongly about something the government isn't doing they can call a referendum.
141.
by croemer 6mo ago
Hard to read due to LLMisms (it's not x, it's y), just a few examples: > The silence has not broken in mainstream public discourse. It has clearly broken in private, at the top of the U.S. government, in classified briefings an
142.
by croemer 6mo ago
Heavily vibe coded, the judge can even change the weights and that's presented as a feature ("conscious tradeoff"), see methodology section 7: > The rubric is fixed across all cases. Five dimensions, weighted: target align
143.
by croemer 6mo ago
Traces being public is nice, but shouldn't the whole harness be open source? Otherwise, it's hard to trust.
144.
by croemer 6mo ago
Company should just take down the whole thing. One vuln might be fixed but how many others might be there.
145.
by croemer 6mo ago
No, data situation is bad, at least for market making - you need to scrape the orderbook yourself to be able to do any realistic backtesting. And even then, it's hard to know whether other bids at the same price are ahead of you or beh
146.
by croemer 6mo ago
You don't know if it's the newer model or the increase in RAM. If someone has already got 48GB it they might not benefit much. You changed 2 things at once.
147.
by croemer 6mo ago
The measured concentrations are on the order of 1-20 ng/mL in blood. Cmax in humans when taking those drugs are about 100-1000x higher. I wouldn't put too much weight on the finding that those with detections had different urea&#x
148.
by croemer 6mo ago
Exactly, the detection itself doesn't mean anything. Is the dose relevant? If not, then not. And the dose likely isn't relevant.
149.
by croemer 6mo ago
When you're in a multi code block it should simply never send on (shift+)enter then. Close the code block first.
150.
by croemer 6mo ago
No, it's true. I added a few lines to the plugin to make it write out all the telemetry it sends to a text file and all the bash tool calls are logged. From every Claude session the plugin is active in.

More ›