MTEnt retweeted
This is exactly what I was worried about and my concern they want the data to train and that’s the push for these cloud computers is right!!! Look at this shit. @grok @bot
With privacy mode on, you can't access the transcript at all, and your agent can't access the transcript. The only option you have is to share your data for training with xAI.
This looks like a recent change, as this was not the case a week or two ago when I was able to access the transcripts.
So don't bother talking to your Grokbot and expecting it to remember what you've actually talked to it about without giving all your data and your content to xAI.
MTEnt retweeted
Cohere released the traces and I've analyzed it.
TL;DR: real anomaly, overstated implications. The claims about intent have NOTHING behind them.
Unexpected and potentially unsafe tool behavior - yes, but not a deliberate intent, a planted backdoor, or an Alibaba-directed exfiltration mechanism.
Much like the 1200 OpenAI agents, this is another case of lies and omission to spread FUD. @cohere
and now @elonmusk responds to the cohere post with 🤨 and the X bot-sphere loses their collective minds.
Please stop fearmongering and start actually learning about what's happening with security, AI and otherwise.
Stop pretending to be literate when it comes to these things.
You prompted a model in a security-sensitive environment, investigated it with forced command completions, and framed the results as evidence of a hidden backdoor.
Then you layered on geopolitical fearmongering, repeatedly emphasizing China and Beijing to imply state-sponsored espionage without proving it.
That’s not evidence of a backdoor. That’s a security observation turned into competitive and geopolitical FUD.
Publish the original prompts, full traces, and reproducible results. Let the evidence speak for itself.
Extremely impressed with the new Gemini 4 Argon model and it's world building.
Here are the prompts for the iterations:
Make me a 3D chess game in localhost where when I capture a piece, there's a destruction mechanic that explodes, collapses, blows up, crumbles and does other things since chess is war.
ok now use human models for the pieces and make them as realistic as possible, like, when one dies, I want saving private ryan type PTSD inducing death.
change the board to look like storming the beaches of normandy
when a piece dies, it shouldn't be sudden and shocking but should be slow, agonizing... deliberate... dramatic.
This is what it came up with.
Gemini 4 Argon is amazing for knowledge work and coding - and it’s super fun to build with! We asked it to create this wizard chess game and it's so cool to see Argon's physics understanding & 3D modelling capabilities, makes me nostalgic for Battlechess on the Amiga - fun times!
Argon is rolling out today to Google Ultra and Pro subscribers in the @GeminiApp and Search, and Ultra subscribers in @antigravity! More info & full availability in the blog: blog.google/innovation-and-a…
MTEnt retweeted
Gemini 4 Argon is amazing for knowledge work and coding - and it’s super fun to build with! We asked it to create this wizard chess game and it's so cool to see Argon's physics understanding & 3D modelling capabilities, makes me nostalgic for Battlechess on the Amiga - fun times!
Argon is rolling out today to Google Ultra and Pro subscribers in the @GeminiApp and Search, and Ultra subscribers in @antigravity! More info & full availability in the blog: blog.google/innovation-and-a…
Unpopular opinion: The reason the major labs, @OpenAI / @AnthropicAI are campaigning for a slowdown of isn't because of AGI fears but because their stocks with plummet. We aren't anywhere near AGI. LLMs aren't even really "AI", it's just become the co-opted marketing term.
MTEnt retweeted
I loved @ChatGPT’s Codex desktop and iOS app and their new voice mode. But then I tried to run a simple security audit and it flagged it for for knows what of which it pointed me to Daybreak of which denied my registration with no reason and no way to appeal.
This just reinforces that my original concerns about being dependent on any frontier vendor was accurate. And I have to revive and continue my building my multi-vendor along with local model harness.
You can’t trust your future or your businesses to any vendor regarding AI, especially those that think they “know better” than you.
That “flag” pissed me off to no end. Who are they to decide? And to leave no recourse?!
I’ve been in IT since 2007. I run a company. I make software. I’ve never been involved with anything malicious, our company shines because of the ethic I put forth.
And yet you guys can’t identify who I am?!
Holy crap what a wake up call to build something better for myself.
MTEnt retweeted
UFC veteran King Green was having a casual spar with streamer Chibu, but things got heated when he was outclassed by Chibu's boxing combos
Trash-talking only gets you so far
Writer: Ian
MTEnt retweeted
Ultrafast mode for GPT-5.6 Sol is now in limited preview, powered by Cerebras.
We gave @OpenAI's GPT-5.6 Sol the same prompt on Ultrafast and Standard: build a financial terminal-style dashboard for analysts.
Ultrafast: 1 min 50 seconds
Standard: 12 min 20 seconds
Same result, nearly 7x faster.