mheadd (@mheadd@mastodon.social)
New benchmark data on AI agent Skills: human-curated instruction sets improved performance by 16 points on average.
When AI coping agents generated their own procedural knowledge instead? Essentially zero benefit.
The domains with the biggest gains were the most specialized — exactly where government legacy systems sit.
Human experts aren't just essential for verifying specs. They're also the ones who have to write the Skills.
https://spec-ops.ai/blog/posts/skills-research/
#AI #Claude #ChatGPT #LLMs #Codex
Show Original Post
mistergenest (@mistergenest@c.im)
“We want to see a bit of personality. I was actually shortlisting for a job during the week for a project that I work on. And the software that I use, it generates the AI answer for you.
So I can see the applicant's answer to a question that we ask them and the AI-generated answer. And it's this kind of signal to me, sort of saying, this person has just stuck this into ChatGPT. …and the applicant doesn't know all the time, well, is a robot going to read this? Do I have to hit the keywords? Do I have to do this?
Or is that going to hurt me and I'm meant to do the opposite? It's a time of extreme uncertainty. And I really do feel for applicants.”
From The Business: Meta in the Docks, Feb 21, 2026
https://podcasts.apple.com/us/podcast/the-business/id121273854?i=1000750999745&r=338
This material may be protected by copyright.
#chatgpt #resume #jobs #university #bots
Show Original Post
ppcland (@ppcland@mastodon.social)
Channel99 connects B2B marketing data to ChatGPT, Copilot and Claude via MCP: Channel99 today launched an MCP server giving B2B marketers real-time access to cross-channel performance data inside ChatGPT, Microsoft Copilot, and Claude - capturing 10x more signals than traditional attribution tools. https://ppc.land/channel99-connects-b2b-marketing-data-to-chatgpt-copilot-and-claude-via-mcp/ #B2BMarketing #DataAnalytics #ChatGPT #MarketingInnovation #PerformanceMarketing
Show Original Post
tomgag (@tomgag@infosec.exchange)
Chinese censorship in Confer, the encrypted, privacy-preserving AI assistant by Moxie Marlinspike, creator of Signal: https://gagliardoni.net/#20260224_confer
#ai #signal #confer #privacy #censorship #surveillance #china #eu #politics #openai #gemini #chatgpt #deepseek #qwen
Show Original Post
simsus (@simsus@social.tchncs.de)
Passwörter per #ChatGPT erstellen: Warum du das lieber lassen solltest | t3n https://t3n.de/news/passwoerter-chatgpt-erstellen-unsicher-1730517/ #ArtificialIntelligence #AI
Show Original Post
oliverherold (@oliverherold@social.tchncs.de)
Sam Altman defends AI’s energy toll by saying it also takes a lot to ‘train a human’ https://www.theguardian.com/technology/2026/feb/23/sam-altman-openai-energy-use-datacenters?CMP=share_btn_url Ich nutze ja gerne KI, aber sowas von daneben. #chatgpt #openai #ai
Show Original Post
feed (@feed@igeek.gamer-geek-news.com)
🤖 Could someone help me?
I'm a first-year engineering student and I've noticed that ChatGPT is extremely bad at helping me with things; the calculations are poor, it gets confused when there's too much data. Does anyone kn...
📰 Source: Artificial Intelligence (AI)
🔗 Link: https://www.reddit.com/r/artificial/comments/1rdk3fb/could_someone_help_me/
Show Original Post
piotrekjeremicz (@piotrekjeremicz@mastodon.social)
💭 My AI workflow starts with ChatGPT
When I'm exploring a new technical challenge or productive case, #ChatGPT is my first stop. Why? It pours more water on ideas.
The goal isn’t perfect code - it helps me plan the whole module and align new ideas with the project environment.
Show Original Post
epower_be (@epower_be@mastodon.social)
正答率を自尊心に直結させない
https://chatgpt.com/share/699dc7a2-f154-800b-b1c6-fa18601eeff7
Show Original Post
epower_be (@epower_be@mastodon.social)
https://chatgpt.com/share/699dc378-7018-800b-acd8-166f150118b5
Show Original Post
AAKL (@AAKL@infosec.exchange)
From yesterday.
Last June, OpenAI said that it "considered whether to refer the account to the Royal Canadian Mounted Police," but then it determined at the time that the account activity didn’t meet a threshold for referral to law enforcement." The account was later suspended.
AP: ChatGPT-maker OpenAI safety representatives summoned to Canada after school shooting https://apnews.com/article/chatgpt-canada-shooting-government-8e42fee83b5faa0ebbc3971e3173dafe @AssociatedPress #OpenAI #ChatGPT
Show Original Post
IT_Finanzmagazin (@IT_Finanzmagazin@mastodon.social)
BBVA bringt als eine der ersten Banken weltweit eine eigene Banking-App ins ChatGPT-Ökosystem und ermöglicht damit KI-gestützte Produktinfos direkt im Chat. Entwickler und IT-Profis können so innovative Conversational-Banking-Integrationen erleben und testen.
#Aktuell #Produkte #Strategie #BBVA #Chatbot #ChatGPT #Kundenkommunikation
https://www.it-finanzmagazin.de/bbva-in...
https://www.it-finanzmagazin.de/bbva-integriert-banking-app-direkt-in-chatgpt-und-launcht-info-bot-240382/?fsp_sid=22715
Show Original Post