Elektrine
Log in Register
Paige Chat Timeline Gallery Friends Email Drive DNS Private DNS Domains VPN Kairo Nerve
Remote

G. Allais

@gallais@mamot.fr
mastodon 4.7.2
  • Open on mamot.fr

Lecturer in CS
#Idris 2 dev, all my toots are proven correct

615 Followers
191 Following
30 Posts
Joined April 06, 2017
www:
https://gallais.github.io
irl:
Glasgow, Scotland
pronouns:
he/him
Open post
G. Allais @gallais@mamot.fr
· 2mo ago
What does it even mean to "have just formalized a complicated paper in a language and not needed to know any of said language"? Would anyone use DeepL and proudly proclaim to "have just written a complicated novel in Chinese without knowing any of it"?! What's our epistemic understanding of "formalisation" here? Uninteresting busywork these programming people do?
32
7
13
0
Open post
G. Allais @gallais@mamot.fr
· 2mo ago
Replying to
@jonmsterling@mathstodon.xyz There's even worse; I am astonished that articles purporting to use deep learning to "enhance" pictures meant to be use in research settings are being published. Or that the research is being funded to being with. What do you mean vibing missing pixels into existence is useful for "monitoring of aquatic ecosystem"?! https://www.nature.com/articles/s41598-026-47888-7t
nature.com
7
1
1
0
Open post
G. Allais @gallais@mamot.fr
· 6mo ago

New rule: if you post a pro-AI message on the TYPES mailing list, you need to declare your conflicts of interest.

11
3
1
0
Open post
G. Allais @gallais@mamot.fr
· 5mo ago

I see ScienceDirect has decided to also jump on the bandwagon no one wants

7
0
5
0
Open post
G. Allais @gallais@mamot.fr
· 5mo ago
Replying to
@whitequark @jonmsterling @mei The OG proof of False in Coq on proofmarket was done by Pierre-Marie Pédrot who redefined False to be True (https://xcancel.com/proofmarket/status/419309366135640065)
xcancel.com
6
1
1
0
Open post
G. Allais @gallais@mamot.fr
· 5mo ago
Replying to
@mc @liamoc @amy In person exams and because writing a significant amount of code on paper under time constraints is rather unpractical and artificial, probably focus on code comprehension, identify corner cases in specifications, find good names for variables, add comments documenting unwritten invariants, etc. Which tbh are probably better measures of whether they have a good mental model of the language and its semantics.
6
4
1
0
Open post
G. Allais @gallais@mamot.fr
· 5mo ago
Replying to
Like, I get it, these policy documents are boring AF to read but... that's literally your job? And instead of summarising the document yourself by taking notes through your (alleged) first read of the full document you ask an LLM to summarise it for you?!
6
0
0
0
Open post
G. Allais @gallais@mamot.fr
· 5mo ago

I wonder whether anyone has studied the effects of the shift from in-person PCs to online discussions.

I expect in-person meant people had to be quick on their feet and there was a bonus for good public speakers whereas online discussions over a week allow you to reword your point even write code snippets to make your arguments clearer.

6
2
1
0
Open post
G. Allais @gallais@mamot.fr
· 6mo ago

Not to toot my own horn but I systematically use the curl-based installer for agda-stdlib these days and it's so convenient. 🥰

https://github.com/agda/agda-stdlib/#automated-installation-currently-experimental

github.com
6
1
1
0
Open post
G. Allais @gallais@mamot.fr
· 5mo ago

Presented without comment:

Today: High earners race ahead on AI as workplace divide widens https://www.ft.com/content/0873e3cb-cb02-4b47-941f-14da74149670

Two days ago: Elite law firm Sullivan & Cromwell admits to AI ‘hallucinations’ https://www.ft.com/content/657d86df-5e0d-4d03-bf0c-cb768a58e758

ft.com
4
0
0
0
Open post
G. Allais @gallais@mamot.fr
· 5mo ago
Replying to
@whitequark @jonmsterling @mei Sure, there were also some due to bugs in the checker. But the fact that the OG one was done by changing the statement rhymes nicely with Jon's point.
4
1
0
0
Open post
G. Allais @gallais@mamot.fr
· 7mo ago

Come join us as an MSP lecturer in sunny Glasgow!

https://www.jobs.ac.uk/job/DQS044/lecturer-in-mathematically-structured-programming-790646

jobs.ac.uk
6
0
32
0
Open post
G. Allais @gallais@mamot.fr
· 5mo ago

```quote
By default, PGFPlots assumes that the columns are separated by white space.
```

https://tex.stackexchange.com/a/251245/26482

I wonder what `.csv` stands for? 🤔

tex.stackexchange.com
3
0
0
0
Open post
G. Allais @gallais@mamot.fr
· 5mo ago
Replying to
@maxsnew It reads like toxic "git gud" gamer chat
3
0
0
0
Open post
G. Allais @gallais@mamot.fr
· 5mo ago

I've only seen of ETAPS what the official account has been posting and if the trend of having tons of talks about AI continues, I'll have even less incentive than I already have to go to conferences.

3
1
2
0
Open post
G. Allais @gallais@mamot.fr
· 5mo ago
Replying to
@JacquesC2@types.pl haha! I'm really annoyed by that Lean trend
2
2
0
0
Open post
G. Allais @gallais@mamot.fr
· 5mo ago
Replying to
@constantine @ncf @de_Jong_Tom @ionchy @andrejbauer I have seen the future 15 years ago: https://download.ocamlcore.org/melt/melt/1.2.0/doc.pdf
download.ocamlcore.org
2
0
3
0
Open post
G. Allais @gallais@mamot.fr
· 5mo ago

I am in a very

```
error: shim_lock protocol not found.
error: you need to load the kernel first.
```

time in my life

2
0
0
0
Open post
G. Allais @gallais@mamot.fr
· 6mo ago
Replying to
@maxsnew We may be talking about different guys but in this instance: someone I have cited more than a few times for their great NbE-related work and its application to termination checking. Hence why I even bothered to have a look to try to see what brought us here.
2
0
0
0
Open post
G. Allais @gallais@mamot.fr
· 5mo ago
Replying to
@jfdm@discuss.systems Of course people would vote differently in a purely proportional system but here it is (given the current state)
1
2
0
0
Open post
G. Allais @gallais@mamot.fr
· 5mo ago
Replying to
@julesh@mathstodon.xyz you just need to execute your code rather than normalising it in the REPL
1
0
0
0
Open post
G. Allais @gallais@mamot.fr
· 5mo ago

I want to see critiques of [big systems] that include the material circumstances of their production. It's probably hard to get to some of the details when the people with the insider knowledge still have vested interests.

1
0
1
0
Open post
G. Allais @gallais@mamot.fr
· 6mo ago
Replying to
I don't care if you're finding ways to measure how famous the subject of a question is, whether you're tracking global success rates for different questions, whether you're building a profile of your users' relative knowledge of various domain-relevant characteristics, whether you're controlling how clustered the proposed answers to an MCQ are. Just do something more innovative than adding a time pressure mechanism please! 😭
0
0
0
0
Open post
G. Allais @gallais@mamot.fr
· 6mo ago

@MonniauxD@social.sciences.re https://leodemoura.github.io/blog/2026-3-16-who-watches-the-provers/

leodemoura.github.io

Who Watches the Provers? — Leonardo de Moura

Leonardo de Moura — Creator of Lean and Z3

0
0
0
0
Open post
G. Allais @gallais@mamot.fr
· 5mo ago
Replying to
@liamoc @mc @amy We have a similar setup (for class tests, not sure about exams) but you can't ask for much (both in terms of quantity and sophistication) to be done over 2h. And cohorts are getting large enough that it starts being complicated to organise (as in needing to block book all the computer labs distributed over 3 different floors)
0
2
0
0
Open post
G. Allais @gallais@mamot.fr
· 5mo ago
Replying to
@jfdm Do we know for a fact that GenAI cannot generate the documentation along with the code?
0
6
0
0
Open post
G. Allais @gallais@mamot.fr
· 5mo ago
Replying to
@jfdm Right, I was thinking of (bi?)weekly marking in labs of a couple of functions with students explaining their code. But it makes the whole process a lot more rigid which is tricky for people with a side job or caring responsibilities.
0
0
0
0
Open post
G. Allais @gallais@mamot.fr
· 5mo ago
Replying to
@jfdm@discuss.systems https://www.bbc.co.uk/news/election/2026/scotland/results has a button to make the total constituencies & regions results appear
bbc.co.uk
0
1
0
0
Back
313k7r1n3
Elektrine

Tor hidden service

elekhj7afj4qnrr4yd3bkzslsyo5jgfxw3orgjkhlcxifueodybyiiad.onion

I2P eepsite

j6b6cyk6gjmepjih7jjadxgxvvf3lzzujljuu2v4biemzpg3naya.b32.i2p

Platform

  • Email
  • Chat
  • Timeline
  • VPN
  • DNS

Company

  • About
  • Contact
  • FAQ
  • Lite (no JS)

Legal

  • Terms of Service
  • Privacy Policy
  • Transparency Report
  • Report Abuse
  • Warrant Canary
  • VPN Policy

Support

  • support@elektrine.com
  • Report Security Issue
Mail client setup IMAP mail.elektrine.com:993 POP3 mail.elektrine.com:995 SMTP mail.elektrine.com:465
© 2026 Elektrine. All rights reserved. Server: 06:17:45 UTC