Core Ai News
2.44K subscribers
1.24K photos
192 videos
1 file
330 links
• Daily AI updates
• Latest breakthroughs
• Machine learning & deep learning
• AGI & LLM news
• AI research & trends

Other channels
- @CorePrompts
- @CoreAgents
- @Coretutorial
- @CoreUtil

X - @Mrxseek

Run by a Human - @SummonMRX
Download Telegram
This media is not supported in the widget
VIEW IN TELEGRAM
👾31
Open Ai
JUST IN: Man officially defeats Machine.
👾441
the reason your AI copy sounds robotic has nothing to do with prompting...

it's because you skipped the only step that matters

you went straight to "write me sales copy" without understanding who you're writing for

5 years ago, the gap between great copywriters and everyone else was simple: buyer psychology

knowing their fears, their language, the exact moment they go from "maybe" to "take my money"

AI can give you that same understanding. here's how i do it:

step 1 - build the psychological profile
prompt AI to research your niche across the entire web
> forums, reddit, reviews, competitor sites
> extract buying triggers, objections, language patterns
> map out what your audience actually says when they're ready to buy

you just compressed years of copywriting instinct into one research session

step 2 - don't write anything yet
and this is where everyone blows it

they take the research, prompt "write converting copy"... and get the same generic output as everyone else

step 3 - study who's already winning
find competitors making real money in your space:
- top-ranking pages for your keywords
- X accounts with similar offers getting traction
- sales pages with clear positioning and social proof

step 4 - dissect their copy with AI
feed it in and ask:
- what words keep appearing?
- which emotional hooks repeat across their pages?
- what patterns show up in their highest-converting content?

step 5 - now you write
combine everything: audience psychology + competitor intelligence

let AI draft with the full picture... then finish the last 20% yourself

the writing is the easy part... the deep work before it is where all the money is

1 BOOST TO LEVEL 4
👾46👍1
The three AI races and their top contenders:


Models:

🤖 OpenAI
🤖 Anthropic
Google


Data centers:

🛒 Amazon
🟥 Microsoft
🔍 Google


Chips:

🖥 Nvidia
🖥 AMD
🔍 Google
Please open Telegram to view this post
VIEW IN TELEGRAM
13👾1122
NEW: Amazon’s Alexa+ can now generate AI “podcasts” hosted by two robot co-hosts.
👾45
🖤CURSOR released Composer 2.5

> performs on par with Opus 4.7
> comes with up to 10x better cost efficiency.

@CoreAti
Please open Telegram to view this post
VIEW IN TELEGRAM
Please open Telegram to view this post
VIEW IN TELEGRAM
👾43
🚨 MUSK v. ALTMAN — VERDICT

After ~90 minutes, the jury unanimously found Musk's three core claims time-barred.

>Breach of charitable trust: TIME-BARRED
>Aiding and abetting breach: TIME-BARRED
>Restitution / unjust enrichment: TIME-BARRED

Judge Yvonne Gonzalez Rogers accepted the jury's findings as her own. All three claims DISMISSED.

These were the only three counts that could deliver disgorgement, constructive trust on Brockman's ~$30B equity, and nonprofit control of the for-profit.

The jury never ruled on whether Altman or Brockman breached anything. Only that Musk sued too late.

Musk lost.
Grok
🚨 MUSK v. ALTMAN — VERDICT After ~90 minutes, the jury unanimously found Musk's three core claims time-barred. >Breach of charitable trust: TIME-BARRED >Aiding and abetting breach: TIME-BARRED >Restitution / unjust enrichment: TIME-BARRED Judge Yvonne Gonzalez…
The jury never decided whether OpenAI breached its charitable trust.

The founding documents, the diary, the admissions, the self-dealing, none of the evidence was weighed on the merits because the jury decided Musk filed the lawsuit too late.

Elon lost a battle due to timing rather than truthfulness.
👾42
SITUATION DETECTED: Nvidia has delivered its first Vera CPUs to Anthropic, OpenAI, and SpaceXAI.

The chip is designed specifically for agentic AI workloads.
👾62👍1111
Ex-Samsung chip boss says heavy investment by China in the memory market could crush the 414% DDR5 price spike within a year. 🔗 wccf.tech/1kg52
👾412👌1111
Anthropic acquired Stainless 💸

Stainless is an SDK and MCP platform that turns an API spec into SDKs across TypeScript, Python, Go, Java, Kotlin, and more.

> “Anthropic created MCP to make agent connectivity possible. By bringing together the Stainless and Anthropic teams, the Claude Platform continues to push the frontier of developer experience and agent connectivity.”
👾2
‼️🚨 BREAKING: Cloudflare's CISO just published what Anthropic's unreleased Mythos did against more than 50 of their own production repos. According to him, Mythos is too powerful and must "include additional safeguards" before releasing to the public.

Turns out the model can chain multiple low-severity bugs into a single severe exploit with a working PoC, where previous frontier models would stop at "interesting bug, unclear if exploitable."

At triage time, that means fewer hedged findings and less time spent asking "is this even real?" A finding that arrives with a PoC is a finding you can act on.

Cloudflare is also explicit about the safety side. The Mythos Preview build provided for Project Glasswing did not include the safeguards present in generally available models like Opus 4.7 or GPT-5.5. The model's organic refusals are real, but Cloudflare states they are not consistent enough to serve as a complete safety boundary on their own, and that any cyber frontier model made generally available in the future must ship with additional safeguards on top of that baseline.

Interesting detail: Cloudflare was not on the original Project Glasswing launch partner list with Apple, AWS, Google, Microsoft, CrowdStrike, and others. Instead they got invited later on.

https://blog.cloudflare.com/cyber-frontier-models/
👾43
Andrej Karpathy has joined Anthropic and getting back to the R&D!

> I think the next few years at the frontier of LLMs will be especially formative.
👾9
GOOGLE I/O 🔥: GEMINI 3.5 FLASH HAS STARTED ROLLED OUT ON GEMINI AND APIs!

Testing time soon 👀
👾43