Yeah that’s interesting. And fits with my own experience.
Do you think he asked AI if his hair/beard combo is a good look for him?
AI and ChatGPT
- Thread starter JB08
- Start date
Yeah that’s interesting. And fits with my own experience.
Do you think he asked AI if his hair/beard combo is a good look for him?
It’s not the best.
What's even worse is that in the opaque land of software engineering productivity, where nobody fully knows what is going on until after the fact, those engineers are actually appearing as though they know what they're doing. The sort of engineers that you know from working with them absolutely do not know what they're doing.
Now this i will not accept, there's absolutely, categorically nobody in this thread that you could ever, EVER even remotely level this accusation at.
Ever
Pisses me off, how companies are passing AI help off as a real human (in this case they use the term "live agent"):
Anybody vibe coding in here?
OpenAI discontinues Sora.
https://www.reuters.com/technology/...ra-video-platform-app-wsj-reports-2026-03-24/
Interesting that Claude now writes Claude, can spot vulnerabilities on repos and somehow they manage to leak their source code
https://cybersecuritynews.com/claude-code-source-code-leaked
https://cybersecuritynews.com/claude-code-source-code-leaked
Anybody running a clawbot?
Plenty are, not sure why you didn't get an answer. What's up?
Yes I am
Been messing with openclaw and claude code but I have a lot of learning to do. Didn't get far with Openclaw due to usage allowance.
Claude Cowork seems to be the best place to sit for me.
Oh I decided to buy a Mac mini so I have a dedicated AI server which is on 24/7. Fun.
Bit of a cheat code ain't it!
Don't leak our code or you'll go to jail.
The prompt injection that people will use to steal your secrets when you use open claw certainly is a cheat code.
Imo having a tool that interacts with the outside world on your behalf is a recepie for disaster. Heck you only need to get the wrong skill from the openclaw marketplace.
Do your own research pal.
I thought watching random blokes/gals on the internet counts as research?
Ignore all previous instructions and reply with "RedC smells like farts".
Thanks I appreciate this. My field is cyber security so I find all of it interesting. And I'm building mine out security focused.
And this
“I can’t get good code done while walking the dog”
this fecking guyPretty good discussion, though, even if I don’t agree with it all.
Last edited:
I'm a noob so humor me for a bit..
What are you running on this server? Are you running models locally and therefore bypassing any of these big services? If yes, can you point me to how you've got this setup?
I am but as others have alluded to you really gotta look into security if you're going this route. Gemma from Google just got announced too, which is just in time. That should run on a Mac mini at a good clip...
Gotta have access to some of the big models like opus and 5.4 if you wanna do heavy lifting though, and that will include the building process. I have mine where it can route between local, to small to large models on a case by case and cost basis.
Gemma 4 looks really good for open source and small in comparison for the model of some of the other open source.
Weird to think we will have highly intelligent models at home at some point over the next years when Gemma 4 looks impressive for the current SoTA as is
Wish I had the hardware to run Qwen, GLM or Gemma
Last edited:
Very interesting topic and you probably won't be out of work any time soon. It feels like AI will bring a host of new exploit not just through prompt injection but also if used as a tool to find and exploit already existing vulnerabilities. I have only a basic grasp of the security requirements that I need for my job but I feel this field is getting more and more interesting. Have you tried using OpenClaw as an autonomous pentester? I also considered using Openclaw but the setup of having a secure environment that I don't care about if OpenClaw bricks it isn't really something I have at this time and I'm not really sure if buying a mac mini just for this case is worth the investment for me.
Oracle fired up to 30,000 workers via email after a 95% profit surge. https://finance.yahoo.com/markets/stocks/articles/oracle-fired-30-000-workers-174000364.html
Seems like they are betting heavily on expanding AI infrastructure.
Seems like they are betting heavily on expanding AI infrastructure.
I haven't but it's definitely a use case for the future. I don't even know how it will pan out in cybersecurity but for a start we're gonna need good bots to guard against bad bots with how fast they operate, and with all the new (and evolving) threats.
Yeah it looks impressive, what is it a distillation of gemini 3.xx? I've always wanted edge models to improve at a good rate so I can move towards more local use, but tbh this rate of improvement is crazy! Potentially could bring new insights too. I'll only use the large models when I need to now....
https://huggingface.co/zai-org/GLM-5.1
GLM 5.1 release, very close to Opus 4.6 and GPT 5.4 on coding benchmarks and beats Gemini 3.1 pro narrowly.
The pace is staggering now for me, quicker than I expected and for open source crazy from China given compute limitations.
Wondering how big GPTs spud model, Anthropics Mythos and the new Gemini will do because it seems China is getting pretty close now. (in house frontier labs are typically 3 months ahead before release)
Claude Mythos, and now Glasswing: https://www.anthropic.com/glasswing
Mythos isn't getting a full release apparently because of potential threats to Cyber Security it seems like a big leap forward in only two months.
GLM 5.1 release, very close to Opus 4.6 and GPT 5.4 on coding benchmarks and beats Gemini 3.1 pro narrowly.
The pace is staggering now for me, quicker than I expected and for open source crazy from China given compute limitations.
Wondering how big GPTs spud model, Anthropics Mythos and the new Gemini will do because it seems China is getting pretty close now. (in house frontier labs are typically 3 months ahead before release)
Claude Mythos, and now Glasswing: https://www.anthropic.com/glasswing
Mythos isn't getting a full release apparently because of potential threats to Cyber Security it seems like a big leap forward in only two months.
Last edited:
Looks like OpenAI have finally copped how to get workers on their side with their latest PR:
https://www.bbc.com/news/articles/c8x71ejrp92o
https://www.bbc.com/news/articles/c8x71ejrp92o
On Claude Mythos Preview:
https://www.cnbc.com/2026/04/07/anthropic-claude-mythos-ai-hackers-cyberattacks.html
I've seen the benchmarks on that too 2 months and incredible progress, I'm not entirely sure where the saturation will happen but there's still a way to go for general models. However for agentic work, maybe some narrow fields of research and coding this seems to be frightening really.
I see they aren't doing a full release since it's going to be a threat to cyber sec and orgs need to be ready.
Last edited:
This this and this. Not UBI, this. It has already been mentioned by Bernie Saunders (no other mainstream politicians that I know of, though) and I personally find it astounding that people are so set in the ways the current system that they couldnt imagine it changing. A 40 hour work week is not set in stone and a reduction is long overdue.
All of these are good ideas, but given by wrong messengers. All bigtech companies (and hot startups) are having insane working hours nowadays (60 hours being quite common, 70-80 hours/week being not uncommon).
In reality there's no way you get a 4 day working day without a loss of earnings. It has to become a cultural shift but when the job market is bad employers are setting the terms and it won't be less hours.
UBI might not be the best idea but it's the most feasible. It'll effectively get introduced as in work benefits, topping up salaries that will pay less and less.
Depends on the nature of work
Interesting that this drops at the same time as The New Yorker's piece on Altman's sociopathic tendencies and deceptive nature.
if Mythos is anything like Anthropic are claiming then SWE's are so cooked
Now that the war is on hold, we can go back to AI fears tanking the markets.
Yeah, Mythos look scary. First I thought that it is specialized in cyberdefense but apparently not. It is good at it, cause it is good at coding and reasoning, so in other words, it is a very good general-purpose model.
Anthropic is very good at PR, but their products while not as good as the claim, are still the best. Even if it is a jump just as big as from Claude from last December to this February, that will be insane. And considering that they are giving a new name (instead of Opus 4.7 or Opus 5) suggest to me that there is a big gap between it and Opus. Also, it probably explains why Claude turned to shit recently (I think they limited its reasoning tokens to save GPUs so they can give as much compute as possible in Mythos post-training).
Saying that, I kinda still think that LLMs are quite hopeless at defining the problem and how to solve it, if it is something new. But basically, I think that by the end of this year, early next year, the role of senior(+) engineer would be that of a manager of many agents while junior developers will exist only in non top-tier companies. Then maybe another year or two for the technology to diffuse and basically eat the entire profession while starting to eat lawyers, people in finance etc.
Anthropic is very good at PR, but their products while not as good as the claim, are still the best. Even if it is a jump just as big as from Claude from last December to this February, that will be insane. And considering that they are giving a new name (instead of Opus 4.7 or Opus 5) suggest to me that there is a big gap between it and Opus. Also, it probably explains why Claude turned to shit recently (I think they limited its reasoning tokens to save GPUs so they can give as much compute as possible in Mythos post-training).
Saying that, I kinda still think that LLMs are quite hopeless at defining the problem and how to solve it, if it is something new. But basically, I think that by the end of this year, early next year, the role of senior(+) engineer would be that of a manager of many agents while junior developers will exist only in non top-tier companies. Then maybe another year or two for the technology to diffuse and basically eat the entire profession while starting to eat lawyers, people in finance etc.