ChatGPT can't write you a good paper, but it can prepare an advanced paper substitute that takes far less effort to make than to debunk, allowing you to LARP as a student uncontested for another week. It has no shame and can lie bigger, faster, and more confidently than you can.

Either the tools are some kind of psycho-hazard, deluding people into thinking they work despite contrary evidence, or students know they don't work but are leaning into the currently ascendant "baffle them with bullshit" approach to life.

I can't imagine engaging with prose of any length and substance from one of these tools, or especially technical writing, and thinking "this is a good, well-structured piece of writing that is consistent with itself and the logic of its topic".

In my attempts to become an "AI-enabled" programmer as a senior engineer, one thing I'm noticing is that you can tell synthesized code from code I wrote by hand because it's almost always terrible. Overconfident, badly communicated, and subtly wrong. The required understanding is absent.

Yet the students apparently believe the tools work as advertised, and act like what they submit is acceptable-quality work, or like people can't tell that the required things are not in their heads after they read a summary.

It seems to me that the people being scammed most here are the students: the tools do not actually allow them to work effectively with or produce compelling writing about ideas they themselves decline to understand.

And a paper submitted with even a single made-up reference is not only a complete failure as a piece of writing, but also an obvious academic dishonesty problem. I'd hope the students would notice that competing methods of cheating are more reliable.

Can't the students tell?

If students come to class with such a poor understanding of the material that they noticeably drag down the discussion, whatever they thought they were using AI to do has palpably failed; shouldn't they notice they're doing poorly in the course?

I think it might be originally part of SAM/BAM and now semi-officially belong to htslib and doi.org/10.1093/giga... maybe?

OK, it was right, `/dev/null` fails `os.path.isfile()` and that turns out to be important. I guess it's worth maybe $15.

It's now gotten this though:

echo "test content" > src/toil/test/cwl/empty_file_for_test.txt

I'm a scientist agentically coding right now and Claude Code has come up with:

Bash command

ls -la /dev/null && file /dev/null
Check /dev/null properties

Do you want to proceed?

I even wrote it a full set of tests and everything. I'm not sure it's worth the $20.

For me it's "Google". "Google" is constantly calling me to inform me that people in various far-flung locales are changing my account phone number and to hang up if that's chill.

About ready to give up on PSTN and join the Calyx Institute calyxinstitute.org/membership/i...

Haonan Wu gives a talk on "A k-mer-based estimator of the substitution rate between repetitive sequences"
www.biorxiv.org/content/10.1...
This work tackles the issue of Mash which ignores repeats in the genome, providing better distance estimation #GI2025

How many bases are in a word?