So an interesting case with two news with a difference of 1 day.
The first news about how Claude is becoming lazier. After a boom of users in Claude, they are screwing down performances for the AI. The bot has started giving increasingly stupid answer.
I've noticed it since last year for popular AIs. Example: Google released Gemini 3. The model was smart. After a couple of months, I noticed the model had worsened compared to how it was then.
After some Antropic released a super-duper AI model with the best performance. And after a couple of months, a similar situation.
https://www.theregister.com/2026/04/06/anthropic_claude_code_dumber_lazier_amd_ai_director/
🔗 Original Post
The first news about how Claude is becoming lazier. After a boom of users in Claude, they are screwing down performances for the AI. The bot has started giving increasingly stupid answer.
I've noticed it since last year for popular AIs. Example: Google released Gemini 3. The model was smart. After a couple of months, I noticed the model had worsened compared to how it was then.
After some Antropic released a super-duper AI model with the best performance. And after a couple of months, a similar situation.
https://www.theregister.com/2026/04/06/anthropic_claude_code_dumber_lazier_amd_ai_director/
🔗 Original Post
Andd today, we have the news about the hyper-duper-super AI model by Anthropic. The AI which can hack everything (and your anus).
Buttttt, access to the AI, we give only our brothers(and the Linux Foundation, so that others think we are kind).
Fairy tales : how the model found a 20-year-old zero-day vulnerability; a chain of exploits in the Linux kernel; etc.
I feel like AI companies need trillions in investment right now. When the pipeline of money dries up, all AI companies will fall, or subscribers will face price increases up to the ceiling to cover the real costs of servers.
https://venturebeat.com/technology/anthropic-says-its-most-powerful-ai-cyber-model-is-too-dangerous-to-release
🔗 Original Post
Buttttt, access to the AI, we give only our brothers(and the Linux Foundation, so that others think we are kind).
Fairy tales : how the model found a 20-year-old zero-day vulnerability; a chain of exploits in the Linux kernel; etc.
I feel like AI companies need trillions in investment right now. When the pipeline of money dries up, all AI companies will fall, or subscribers will face price increases up to the ceiling to cover the real costs of servers.
https://venturebeat.com/technology/anthropic-says-its-most-powerful-ai-cyber-model-is-too-dangerous-to-release
🔗 Original Post
We have been living in AGI for 3 years.
https://medium.com/better-programming/i-used-gpt-3-to-find-213-security-vulnerabilities-in-a-single-codebase-cc3870ba9411
🔗 Original Post
https://medium.com/better-programming/i-used-gpt-3-to-find-213-security-vulnerabilities-in-a-single-codebase-cc3870ba9411
🔗 Original Post
hahahahah, how I missed this news
OpenAI (Slop) is creating a code-hosting platform because GitHub's service has been unstable.
They can't study their model because GitHub has a server from potatoes. So, they want to learn from your code without brokers.
https://www.reuters.com/business/openai-is-developing-alternative-microsofts-github-information-reports-2026-03-03/
🔗 Original Post
OpenAI (Slop) is creating a code-hosting platform because GitHub's service has been unstable.
They can't study their model because GitHub has a server from potatoes. So, they want to learn from your code without brokers.
https://www.reuters.com/business/openai-is-developing-alternative-microsofts-github-information-reports-2026-03-03/
🔗 Original Post
Forwarded from Stefan's log (Штефан)
the interesting thing about service provision in the paid model
they require monthly charging for model, after a couple of months, the model has downgraded. but the cost hasn't changed
https://www.bridgebench.ai/hallucination
🔗 Original Post
they require monthly charging for model, after a couple of months, the model has downgraded. but the cost hasn't changed
https://www.bridgebench.ai/hallucination
🔗 Original Post
hahahah, from 1 June, Copilot is increasing the multiplier of models
we are going to the real price on AI
https://docs.github.com/en/copilot/reference/copilot-billing/models-and-pricing#model-multipliers-for-annual-copilot-pro-and-copilot-pro-subscribers
🔗 Original Post
we are going to the real price on AI
https://docs.github.com/en/copilot/reference/copilot-billing/models-and-pricing#model-multipliers-for-annual-copilot-pro-and-copilot-pro-subscribers
🔗 Original Post
let's abuse AI models. for 1-2 years, they will be more expensive
and free models will be more stupid
🔗 Original Post
and free models will be more stupid
🔗 Original Post
gts.crawlcast.one
attanix🐜🍃 (@attanix@gts.crawlcast.one)
let's abuse AI models. for 1-2 years, they will be more expensive
and free models will be more stupid
and free models will be more stupid