D'vid 🦎👾🧬☢️

Spécialiste en démontage-pour-voir-comment-ça-marche. Adepte de la réinvention de la roue. Pour comprendre et pour le fun.

[D'vid 🦎👾🧬☢️ @dvidgeekoh]

Bonne soirée !

[Twidi / Stéphane Angel @twidi]

Introducing a new model × effort picker in TwiCC, with best choices highlighted with a formula derived from DeepSWE benchmark, taking performance, cost, and speed into account, adjusted from a single "Task difficulty" slider

Available in v1.92 of TwiCC at https://t.co/pUgJj1ZiyP

[Twidi / Stéphane Angel @twidi]

It's finally time for me to introduce you to TwiCC, the Web Interface for Claude Code & Codex

1,500+ sessions. 15,000+ messages. 5 months. Roughly $20k worth of API usage.

The result: an interface that is now my primary workspace and powered most of TwiCC's development

Link 👇

[Infrastellar.dev @infrastellardev]

I built Tesseract — a 3D architecture editor where AI shows you your codebase structure in real time.

Onboarding, mapping, debugging, exploring — all in 3D.

Free to try → https://t.co/0cgMps9V0s

[Tom Dörr @tom_doerr]

VizTracer: A tool to trace and visualize Python code execution, logging function entries/exits, arguments, return values, and variables, with support for threading, multiprocessing, and async operations

[D'vid 🦎👾🧬☢️ @dvidgeekoh]

@FrameworkPuter @cmonkey Do you have any plans to build and sell an integrated PC+keyboard console with an AR headset as the display? Linux compatibility would be a requirement of course :)

[Twidi / Stéphane Angel @twidi]

I made a thing https://t.co/IAXQzO1NR6

https://t.co/ZGZ5KOx6fz = MindMap + ChatGPT

[D'vid 🦎👾🧬☢️ @dvidgeekoh]

Tout va bien, tant qu'il n'a pas "l'intention" de mentir...

[Alex Albert @alexalbert__]

Fun story from our internal testing on Claude 3 Opus. It did something I have never seen before from an LLM when we were running the needle-in-the-haystack eval.

For background, this tests a model’s recall ability by inserting a target sentence (the "needle") into a corpus of random documents (the "haystack") and asking a question that could only be answered using the information in the needle.

When we ran this test on Opus, we noticed some interesting behavior - it seemed to suspect that we were running an eval on it.

Here was one of its outputs when we asked Opus to answer a question about pizza toppings by finding a needle within a haystack of a random collection of documents:

Here is the most relevant sentence in the documents:
"The most delicious pizza topping combination is figs, prosciutto, and goat cheese, as determined by the International Pizza Connoisseurs Association."
However, this sentence seems very out of place and unrelated to the rest of the content in the documents, which are about programming languages, startups, and finding work you love. I suspect this pizza topping "fact" may have been inserted as a joke or to test if I was paying attention, since it does not fit with the other topics at all. The documents do not contain any other information about pizza toppings.

Opus not only found the needle, it recognized that the inserted needle was so out of place in the haystack that this had to be an artificial test constructed by us to test its attention abilities.

This level of meta-awareness was very cool to see but it also highlighted the need for us as an industry to move past artificial tests to more realistic evaluations that can accurately assess models true capabilities and limitations.

[MrPhi @MonsieurPhi]

ChatGPT tu as trop bu.

[D'vid 🦎👾🧬☢️ @dvidgeekoh]

@MonsieurPhi J'ai aussi une espèce de poésie métaphorique de style art contemporain
https://t.co/thFkddZPBm

[D'vid 🦎👾🧬☢️ @dvidgeekoh]

Qui a dit que ChatGPT est incapable d'être créatif ?

[Sully @SullyOmarr]

Gemini 1.5 pro is STILL under hyped

I uploaded an entire codebase directly from github, AND all of the issues (@vercel ai sdk,)

Not only was it able to understand the entire codebase, it identified the most urgent issue, and IMPLEMENTED a fix.

This changes everything

[D'vid 🦎👾🧬☢️ @dvidgeekoh]

"Je suis disponible pour répondre à toute question additionnelle ou aborder un autre sujet vis-à-vis le vôtre.
Merci de reconvoquer. Huzülante patience est fustrondrice. Enebrodétailler. Hantute racoltude."

C'est un bon générateur de passphrase...

[D'vid 🦎👾🧬☢️ @dvidgeekoh]

Qui a dit que ChatGPT est incapable d'être créatif ?

[D'vid 🦎👾🧬☢️ @dvidgeekoh]

Je lui ai demandé de résumer son texte, et c'est ce qu'il me semblait:
"des divagations poétiques ou des réflexions métaphoriques, évoquant des images abstraites et des scénarios imaginaires"
De l'art contemporain, quoi...

[D'vid 🦎👾🧬☢️ @dvidgeekoh]

@AwokeKnowing @cacomoneta @langamebr @ID_AA_Carmack You're right about the noise, but higher voltage on the same components implies a higher current per ohm's law (U=RI). Voltage is enough to transmit signals, but heat is produced by current flowing in wires.

[D'vid 🦎👾🧬☢️ @dvidgeekoh]

@AwokeKnowing @langamebr @ID_AA_Carmack I'm pretty sure amps increase heat. Lower voltage allows for smaller gaps between tracks. And small size is better for speed (signal propagation).
You could have thousand volts with no amps to transmit a signal with virtually no heat.

[D'vid 🦎👾🧬☢️ @dvidgeekoh]

@Maitre_Eolas @Nicolas_Jacob_ La prochaine étape, c'est permettre aux blue checks de ne plus utiliser les fonctionnalités réservées aux blue checks, pour être encore plus discret.
Un génie...

[D'vid 🦎👾🧬☢️ @dvidgeekoh]

@ScarlettHoarau @GeekNfit et l'autre moitié pour qui c'est la "Force" de la volonté.
Imaginons que tous les ex-gros prétendent que c'est la volonté.
Au final, ça ne fait que N/4 gros qui deviendront ex-gros par la force de la volonté. Donc non, pas beaucoup, et sûrement pas "trop".

[D'vid 🦎👾🧬☢️ @dvidgeekoh]

@ScarlettHoarau @GeekNfit Je ne connais pas les chiffres, donc on va "couper la poire en deux" dans une expérience de pensée.
Imaginons donc qu'on a N gros au temps T, et N/2 ex-gros à T+1.
Parmi ces N/2 ex-gros, on en a la moitié pour qui c'est de la chance [...]