“nearly 66% efficacy” - so “somewhat bug-free” then…
Im not taking the piss - seems a major - if over-advertised step forward. 
Yeah - the original press release has a less hypey headline with “bring the dream of bug-free software one step closer”
The 67% is better than it sounds. It means that 67% of the time it comes up with proven bug free software.
So you know whether there are bugs or if you have proven bug free code. You just don’t get the latter one third of the time.
I have been working with agentic AI in the last week or so.Autogen and tonight crew-AI (v promising). There is also taskwarrior.
I used llmstudio for autogen and created pipelines for various tasks. So coder, reviewer, QA, CTO, CEO and marketing etc.
There we can have an LLM take a spec, write out a detailed spec, another writes the code, another runs the code (in docker) and repeat till it’s working. Then QA may add test requirements and back around again. The CTO then may add/remove features and back around agian.
So what this does is allows planning and also testing of code. No more getting errors and fixing those, the agentic network does that.
So this was all done using codellama, mistral etc.
With taskwarrior and the most exciting one crewai (as it allows delegation and very large teams (crews) I can use ollama to serve many local llms.
Here’s an example
llama2-uncensored as CEO
codellama → Developer
deepseek-coder → QA
phind-codellama → CTO
and so on.
So we can have a delegatable agentic network using the best local LLMs for each persona.
We can even have
samantha-mistral → for wee feelgood chats wile the rest are working.
What I am poking at here is not a single AGI / AI model approach and also purely 100% local and private approach that has memory, multi agents, multi models and RAG etc. capabilities. These can use tools to scrape the web or calculate etc. with precision.
So a toolkit of privacy first, best of breed per domain local models.
It’s quite interesting and will make yer eyes pop out. Early early doors yet, but I think to extend, clarify and add planning and collaboration this setup holds great promise.
Of course, end goal is using SAFE to make this actually SAFE and private, but that part I suspect will be much simpler than it sounds.
A few demos on YouTube could be an interesting way of getting new eyes on Safe Network the run up to launch (whether the stuff works [cough] or not 
).
Ok added
meditron → Doctor, just in case 
Tell me, Mr. Irvine, if I can count on a launch before my daughter’s birthday… I would like to take a sabbatical year and travel with her “freely” and with some exchanged maid at a “good price”
I don’t know when your daughter’s birthday is 
I think in the next few weeks you won’t see launch, but you might be quite surprised all the same.
I was too slow there, here’s your first good news MaidSafeCoin (EMAID) now available on the BitMart exchange!
We’re not finished yet though, much bigger, not exchange related news likely soon. The wheels are turning.
Here we go again, is @scottefc86 excluded now he already has a podium?
I’ll put more thought in this time before I waste my ticket again!!
That word … 
Trying to stifle the competition 
I heard if I guess another one right I get free SafeNetwork uploads for life 
Alright, you first this time!! 
Lets hear it.
I’ll go with some sort of collaboration/partnership. Can’t predict the tech cos I ain’t got a fudgin clue what’s going on
We are kind of on the same page.
Going with funding secured idk may be as crap as my last guess 
.
Not to be confused with deploying more capital
I’m sure we’ll found out SOON
A partnership was all I came up with but I’m not convinced.
Easier to say what it isn’t likely to be I think, but that’s an infinite list.
Ayrshire planespotters report a Gulfstream G650 registered to Tesla Corp has been parked at PIK since early this morning… #JustSayin
It’s not a VC
With the stable test nets I am sure other entities have noticed and are now bashing the door down to become partners.