Claud Fable 5 - @AnthropicAI@claudeai
I’ll admit I am impressed… I’m using it as a mentor in my Agentic System Development Freelancing/consultation service.
It was really detailed in strategy and provided good proof. Tip is to be ultra detailed about what you want. This model is capable of doing it!
Introducing Claude Fable 5: a Mythos-class model that we’ve made safe for general use.
Its capabilities exceed those of any model we’ve ever made generally available.
AI companies are running into dead ends on data to train their models with…
Scraping everything.
Seems the next thing to train it on would be straight up human behavior and feeling. But, to me that sounds like the part where AI might learn different types of emotion.
Who knows - just a thought.
@pcshipp Depends on how you think but most of the time start with backend and build in small layers. Look up Agile, lol. But if you’re a reverse engineer style thinker, maybe front end doesn’t hurt to see first and then just go through.
@markproduct Founders are visionary’s and people who see the problem in deeper levels and “connect each dot” to get solutions. AI slowly becomes a game changer for minds like that by simply building what the mind of a founder imagines…
https://t.co/vUah6jPRxg - I’m actively rewriting the way students learn. Simply by looking at class notes/slideshows/webpages you can generate a * study guide in a SINGLE button click * - I’m serious. AutoStudyAI uses analyzation to learn how you learn in different patterns and utilizes AI Tutors. AutoStudyAI never searches the answer on the internet and all material is straight from your class - no hallucination or guessing. Check it out!
I can see that as a challenge for sure! I’m simply testing the handoffs by building agents that utilize different protocols on the application layer - analyzing the packet from the agent to the server (dest.) - and finding the differences. I’m not extremely familiar on all of it but it’s exciting to see the merge between AI & IT!
I’m surprised we haven’t seen an Application Layer protocol yet for agent to agent/agent to server requests… in terms of prompt injection precautions? Basically the flow of data from the agent to its servers (Anthropic / OpenAI / Google / …etc)
If I had to be delusional and give a guess - It's the same thing when you use a Galton board and the random number theory and how, even with randomness, it still finds its way to the middle. And I guess telling a data model to do something 'random' would be interesting + temperature...
If you aren’t into investing… you need to be.
Specifically, the tech industry is on its way to become an extremely profitable market long term. It’s extremely volatile at the moment - but thinking year 2040-2050 even…
We really are about to enter the future.
Has anyone else gone from IT to learning programming?
I feel like I see code so differently than what it should be because my ideals are based off a systems-thinking standpoint. I’m seeing the code I’m building as a network - not really a project…
@futureispsycho@0xdoug Sure, you can’t hire your way out of a bad pretrain… but you can hire your way out of bad evals, bad RLHF, bad scaffolding, and bad product — which is where, I would think, around 90% of the ‘issues’ actually are. Labs keep funding GPUs and starving the layer that ships.