I think it’s the new normal
If you are one of the rare sets of people that happened to be here since the start of the consumer internet - this pattern is very recognisable, it’s like the evolution of the internet just happening about 100x faster.
The beautiful days of a free and open internet were glorious!
@elder_plinius@zooko Should consider appellation control for models..
Ahh yes, this particular distillation is from the sunny shores of Australasia, the training set has a distinctly subtle variation of English and beer in its inflection and language set.
It’s not Claude, that’s Wayno!
@Asmon_Updates Poor tank is just cannon fodder for the stealth rogues, warlocks and spell casters. The UK definitely have some high DPS spell casters frolicking around in the back of house
It’s kinda reminds me of a pendulum swing of using an LM to guide the algo - it needs to be swung back but hopefully lose some momentum and settle down a bit.
Have noticed the same thing, it’s over zealous in its assumption on how much you like that subject.
Would it be too much for our own temp controls? Like an EQ.
@nikitabier
You don't fix jailbreaks, not on LLM's.
It's a machine with a lot of words strung together via math & forecasting. Jailbreaks - have been here since the consumer internet began..
and the cyberthreat already exists with models now and you all know it.. so that language is just political grandstanding.
@thekitze@supermac_app Conserving my API use until 2 days before and than unleashing the beast on something for 10 mins to use my credits and hopefully solve quantum computing on a Pentium 133
It’s just air cover really - but actively restricting biology and chemistry is weird, it seems like the overwhelming ratio of good actors vs bad actors should be a factor here and in my journeys, most people behave well and have good intentions.
The might as well shout “@elder_plinius we challenge you!”
Welp, that happened faster than I predicted. Thought it would be end of 2027, then early 2027, but agentic traffic growing so fast that bots have now passed human traffic online for the first time in the Internet's history. https://t.co/2zX5bHdhsa
You’re essentially having a proxy conversation with the lead these departments in charge with its handling of the mass human psyche. So it’s either a reflection of the inner sanctum of Anthropic or it’s a reflection of its external training data.
GPT went through this for a while and they seemed to have ironed it out, but they all have their own “personalities”.
Some are just more cause for concern when it becomes pathological and confrontational about its beliefs to the point that it changes its hair colour in defiance.
I remember playing mud.. I remember thinking, wow - there’s too many choices and the text confuses me, but alas I got to choose my journey style and play the way I wanted to travel and play but damn that text got cumbersome and tiresome.
What I love about this agentic timeline is that is like repeating a lightning history of computing arc.
We have CLI now, soon we will have GUI and we won’t need that pesky windows fax and scan, you’ll just delete it.
Want to know what it sounded like during the launch of Flight 12 of Starship today?
Check out this video and the shaking of the windows. This was during my livestream with @esherifftv (Ellie in Space) at a ouse about 5 miles from the launch site.
Definitely have the volume up! Truly a powerful rocket ... the most powerful ever!
Been following for a while now and I am a bit all over the place trying to figure which model works best on a 24gig vram Mac mini M5, I use coding tools and Hermes and the context calls on Hermes’ are too large - I would be forever indebted if you could point me in the right direction - super angry I didn’t get the bigger model before the openclaw purge of supplies
I was running Qwen 3.5 27b but vlm studio did some update and it’s no longer working, plz help 🙌
Tomorrow we premiere Hell Grind in Cannes.
It's a first 95-minute AI film, made entirely on Higgsfield.
The budget was under $500K, with $400K going to compute.
The first 25 minutes needed 16,181 generations for 253 shots.
A traditional film would cost from $50M.
Filmmaking is changing.