Andy Pilate
443 posts
Andy Pilate
@Cubox_
Internet was a mistake. I hate emojis.
Paris Katılım Şubat 2011
86 Takip Edilen498 Takipçiler
Quelqu'un aurait ici un abonnement ChatGPT Pro X5 (le truc à 103€) ? Pour une bête question d'usage (juste voir si vous avez bien la fonction chatgpt.com/skills)
Thanks !
Français
@eiranoirx For context, at my school we had to recode malloc entirely from scratch using mmap. While it might not be what the actual malloc does, you kind of have to get into it. I'll share the project info here in a bit.
English
malloc doesn't actually give you memory. it gives you a promise.
the OS hands over real RAM only when you touch the page. so you can malloc(1GB) on a 512MB machine and it won't fail immediately.
it fails later. with a SIGKILL. no warning.
linux has a process called the OOM killer that walks through everything running and decides who dies. it scores each process and just kills the heaviest one.
you can protect your process from it.
-1000 means never kill this. +1000 means kill this first. PostgreSQL and Redis do this in prod.
so next time malloc returns NULL it's not the OS running out of memory. it's the OS finally admitting it lied to you.
English
#VisaSécurité | 🔁 Mise à jour du référentiel PACS : une version 2.0 permettant d’adresser un panel plus important de bénéficiaires pour des prestations mieux adaptées à leurs besoins.
Plus d'informations sur :
🔗 cyber.gouv.fr/actualites/mis…
Français
@psmyrdek @OpenAIDevs I think I saw on the change log that they fixed that today
English
Codex not stopping mid-task when limits are gone is really something worth highlighting. @OpenAIDevs
English
@OrenMe @code @GitHubCopilot In a few days they switch to token based billing so that's not really gonna be useful anymore
English
Finally some clarity around session limits for @code
In the new @GitHubCopilot desktop app you can see the session and weekly limits
Will probably land in @code soon
We see this is happening in all tools that haven't done so yet(think I saw Gemini added rate limits also)
I wonder what we will start seeing once AI provider companies start turning profit - maybe we will get next frog leap soon
Got more questions?
docs.github.com/en/copilot/con…
English
@0xSero Any way you could get an estimate of how much percents is how many tokens, for a given model?
English
I've been spamming 2-4 sessions on Opus-4.7 (Low) via the Droid Max sub for the last 24 hours and I am honestly having an amazing time.
The harness is very stable with the model, thinking is just a waste of money, and tons of usage.
- max 200$ plan
Disclaimer: I'm Sponsored by Factory but they don't tell me what to post, this is my honest experience. I think this + Codex 200$ is everything you'd need (for now)
English
Auto model selection in VS Code now picks the best Copilot model based on task complexity, availability, and health for efficient, reliable coding assistance.
• See which model is used by hovering over responses
github.blog/changelog/2026…
English
@patchli_fr J'étais persuadé que c'était sur liste, vu le terme que je n'ai vu employé que dans ce cadre. Je retire mes conneries alors
Français
Fun fact, mon daron le giga transphobe qui me lâche que c’est pas normal que j’ai droit à une ALD et lui non
patch@patchli_fr
La putain de bordel de sa mère j’ai mon ALD pour la transition
Français
@payloadartist If that's even gpt5.5 and not just an open source model lying
English
It's not worth at any cost to send private data to shady vendors selling OpenAI access for cheap.
aditya@adxtyahq
Chinese students are buying GPT-5.4/5.5 and Claude API access from Xianyu/Taobao proxy sellers for almost 96-97% cheaper People are apparently burning 100M+ tokens a day for like $1 and vibecoding nonstop.
English
Follow-up on the ChatGPT Business promo:
Officially, Business should get the same per-seat Codex limits as Plus.
But in my replies, several users say their Business seats burn through the 5-hour Codex window much faster than Plus. Some say it feels like less than 1/4.
I’ll test it myself and share the actual usage numbers.
So the real question is:
did we find a deal, or did the deal find us?
Kai@hqmank
Quick update: 🎁 ChatGPT Business 48-month promo seems to vary by region. For 2 seats: UK: £11/mo total AU: AU$25/mo total US: $20/mo total Per account: UK: £5.50/mo AU: AU$12.50/mo US: $10/mo If you use Codex heavily, this is still one of the best deals around. links in reply 👇
English
@hqmank @OpenAI @OpenAIDevs Any updates on this? Any good numbers? I'd be curious if someone can test and prove this happens between a "normal" buisness account and a "discount" buisness account and a plus account.
English
Could the @OpenAI @OpenAIDevs team clarify how usage limits work for the Business/Team plan when using the recent promo code?
Many people are reporting that the limit runs out very quickly, even with Fast mode turned off and lower models selected.
Maybe we’re missing something in the setup, just trying to understand the expected behavior.
Anyone else seeing this? Reply below 👇
Kai@hqmank
Follow-up on the ChatGPT Business promo: Officially, Business should get the same per-seat Codex limits as Plus. But in my replies, several users say their Business seats burn through the 5-hour Codex window much faster than Plus. Some say it feels like less than 1/4. I’ll test it myself and share the actual usage numbers. So the real question is: did we find a deal, or did the deal find us?
English
Andy Pilate retweetledi
Preliminary investigations conducted by the Chinese authorities have found no clear evidence of human-to-human transmission of the novel #coronavirus (2019-nCoV) identified in #Wuhan, #China🇨🇳.
English
@CaddyComa @Cubox_ @notnullptr so when Bruno becomes enshittified, I can just move to another one, because the lock in is honestly quite low for me
English
sorry, why are people paying for a http client? and a bad one at that?
English
Peut-être un peu tard pour patcher mes VM de la CVE Copy Fail
Français
@burkeholland We didn't have to update the whole thing when installing a single package
English
