Tootfinder

response can be inconsistent

hmm, discussing the discrepancy between traditional printers points, 72.27 per inch, and computer points, a round 72 per inch https://lobste.rs/s/xsifwf/points_are_weird_inconsistent_unit
it occurs to me to wonder how much influence older low-fi computer printing tec…

Points are a weird and inconsistent unit of measure
10 comments

Are you in tech and outraged about generative AI? Is it being forced down your throat at work?
Here's a nice vindictive way to get a little revenge if you want:
1. Find a project that contains slop code.
2. Optionally, identify specific files or functions that are LLM-generated. I guarantee you that on average, this code has not been adequately tested/inspected, even/especially if it contains LLM-generated test cases.
3. Make up a reason the code could be flawed, bonus points if it's subtle or hard to test. Don't put effort into this or try to actually find a flaw. Just make something up at random.
4. Report your made-up defect as a bug.
That's it. If anyone ever questions you on the incorrect report, just say "oh I used an LLM and it said there was a bug so I reported it." (Don't actually use an LLM, that would be feeding the bubble.)
Note that you are showing the creator of the code the exact same amount of disrespect that they've shown you by publishing slopcode in the first place. I'd bet odds are 50:50 or better that if a human actually follows up on the report, even though they'll find out that the bug report is wrong, they'll find and fix some other subtle flaw in the LLM-generated code, so this is actually helpful in a way.
For step 3, try to get creative. Like "logic in decideUVParameters can cause state to be inconsistent in some cases." If asked for a steps to reproduce, either make one up if it's easy to do so, or say "I forgot how I triggered this." Surely they can ask an LLM to figure out conditions that would trigger the bug ;).
#AI #LLMs #GenAI

Not a foul on Jones and absolutely inconsistent to how the game has been officiated to this point. Darren England is constantly in the way. Ridiculous.
#LFC

@… ahaha I have noticed this but in my experience it seems inconsistent!

In the article "Don't put aria-label on generic elements like divs", @… explains that the ARIA specification prohibits using aria-label this way, and then tests with different web browsers and screen readers.
Conclusion: Because there is no defined behavior, different software handle the aria-label inconsistently. Users may not have the experience you design…

Congrats to AI on finally finding a consistent, repeatable, reliable way to make computers inconsistent, non-repeatable, and unreliable

Been getting these random #npm issues lately 😕
Scenario: dependency has optional peer dep that installs pre-built binaries based on current OS. I npm install it on macOS, it gets the macOS binaries & put it as non-optional dep in package-lock.json. CI runs on Linux, it got confused & failed installation. And npm ci doesn't skip incompatible peer deps.
Relevant issues:
- …

Where You Tap Matters: A Probe-and-Model Benchmark for Open-Set RF Fingerprinting
Gabriele Oligeri, Savio Sciancalepore, Ingrid Huso, Fatima Al-Mousawi
https://arxiv.org/abs/2607.21564 https://arxiv.org/pdf/2607.21564 https://arxiv.org/html/2607.21564
arXiv:2607.21564v1 Announce Type: new
Abstract: Radio Frequency Fingerprint Identification (RFFI) enables transmitter identification at the physical layer by learning device-specific impairments from received signals, yet the literature is inconsistent about where in the receiver chain those samples should be collected. Since distinct transformations are applied to the signal by the different receiver operations, i.e., carrier recovery, gain normalization, pulse shaping, and timing recovery, they can either tighten within-transmitter variability or suppress the features RFFI requires for classification. We present a systematic real-world evaluation of open-set, reconstruction-error RFFI using data collected at five probe points along a standard BPSK receiver chain. Our results show that RFFI is strongly probe-dependent: timing recovery and, to a lesser extent, carrier recovery enable low false-acceptance operation with limited in-distribution-out-of-distribution overlap, whereas other stages often require a false-acceptance ratio above 0.1 to achieve a true-acceptance ratio of 0.9. To test the validity of our findings across model selection, we benchmark several LLM-designed autoencoders using a controlled pipeline that holds preprocessing and MSE scoring fixed. These architectures confirm that RFFI is probe-dependent. Moreover, they do not outperform the baseline at the chosen operating point and typically increase training time. Overall, probe selection dominates reconstruction-based open-set RFFI performance, more than the autoencoder complexity.
toXiv_bot_toot

Part of the ability to detect bot-generated content is simply *good taste*. Humans maybe cannot describe what is “off” about the crud, yet still can see that it’s off.
Maybe it’s inconsistent perspective lines. Maybe it’s impossible focal aberrations. Maybe it’s the odd number of teeth. The redundant sentences. Excessive adjective use. Those who have developed good taste in text or visual art organically may not be aware of the specific technical issues, yet still recognize slop as slo…