post (@post@sopuli.xyz)
gpt-5 leaked system prompt
https://sopuli.xyz/post/31698962
Show Original Post
post (@post@sopuli.xyz)
‘I Feel Like I’m Going Crazy’: ChatGPT Fuels Delusional Spirals
https://sopuli.xyz/post/31698728
Show Original Post
gadgetbond (@gadgetbond@mastodon.social)
GPT-5 integration in Microsoft Copilot brings smarter AI assistance, improved reasoning, and better context handling for productivity, coding, and enterprise workflows. #GPT5 #Microsoft #OpenAI #GitHub #Azure #ChatGPT
https://gadgetbond.com/microsoft-copilot-gpt-5-ai-model-integration/
Show Original Post
microsoft-copilot-gpt-5-ai-model-integration (@microsoft-copilot-gpt-5-ai-model-integration@gadgetbond.com)
Microsoft adds GPT-5 to Copilot with new smart mode for better AI responses
Microsoft’s new smart mode in Copilot uses GPT-5 to automatically select the best AI approach for quick answers or complex multi-step tasks across platforms.
https://gadgetbond.com/microsoft-copilot-gpt-5-ai-model-integration/
Show Original Post
Data_Threads (@Data_Threads@piaille.fr)
🇳🇱 Wikipedia en néerlandais :
1. Jonathan Holslag #JonathanHolslag
2. Aldo van den Nieuwelaar #AldoVanDenNieuwelaar
3. ChatGPT #Chatgpt
Show Original Post
Data_Threads (@Data_Threads@piaille.fr)
🇵🇹 🇧🇷 Wikipédia en portugais:
1. ChatGPT #Chatgpt
2. Hugo Motta #HugoMotta
3. Jorge Costa #JorgeCosta
Show Original Post
lowyat (@lowyat@mastodon.social)
OpenAI Releases GPT-5 Model For All ChatGPT Users #ai #artificalintelligence #chatgpt #gpt5 #openai
https://www.lowyat.net/2025/361921/openai-releases-gpt-5-model-for-all-chatgpt-users/
Show Original Post
bibliolater (@bibliolater@qoto.org)
Thank you for the recommendations; however, I think I will give it a miss.
#AI #ArtificialIntelligence #ChatGPT #Technology #Tech
Show Original Post
RaphaelWimmer (@RaphaelWimmer@hci.social)
@Sobex And it's highly reproducible (but you get different ridiculous explanations each time). Funny that they do not have a set of regression tests that they run for every model.
Show Original Post
SEOSuedwest (@SEOSuedwest@mastodon.social)
GPT-5 already rolled out for 20 percent of paid users:
https://www.seo-suedwest.de/9906-gpt-5-schon-fuer-viele-nutzer-in-chatgpt-verfuegbar.html
Show Original Post
display (@display@friendica.helvetet.eu)
Open AI, bolaget bakom Chat GPT, har släppt en ny version av sitt ai-verktyg. Den marknadsförs som snabbare och smartare än sina föregångare – klok som en doktor enligt Open AI.#ai #openai #chatgpt #tech
Nu är GPT-5 här – Open AI: ”Som att prata med en expert på doktorsnivå”
Show Original Post
stovis (@stovis@fosstodon.org)
Many people seem unaware of the limitations of LLMs. So I'm building a "library" of examples that demonstrate this so people can try for themselves.
For example "You can't lick a badger twice" https://bsky.app/profile/did:plc:knlwbyvg36keh64h6awokbqx/post/3lnhxkdywzc2m.
Or eating rocks, putting glue on pizza https://www.bbc.com/news/articles/cd11gzejgz4o
Or "How many times does letter X appear in this word"
Any others? Please share!
Show Original Post