Hello ,
Valheim has finally reached version 1.0.
This is excellent news for the gaming industry and unfortunate news for my productivity.
The Deep North is the game's final biome, which means there is one frozen corner of Valheim left to conquer. Leaving it unconquered would be irresponsible.
Possibly cowardly.
Fortunately, I have never allowed questionable time-management decisions to interfere with questionable tactical decisions.
Between writing, the day job, and marauding through the frozen wilderness with weapons of uncertain quality, it has been a busy week.
On the Writing Front - The Jags of Jotunheim
I submitted another short story to Chris Kennedy Publishing for the Four Horsemen 2.0 this week.
The Jags of Jotunheim features CASPers, intelligent jaguars, and MinSha fighting on a death world where the wildlife is nearly as dangerous as the enemy.
The planet wants everyone dead.
The MinSha are helping.
The story was written for the new Four Horsemen Universe 2.0. One of the things I enjoy about that universe is that it can support serious military science fiction while still leaving room for enormous combat armor, alien mercenaries, and intelligent jaguars with strong opinions about tactics.
This follows my recent submission of Last Stand of the Werewolves for CKP's Talking with Strangers anthology.
With both stories out the door, I can now focus on Book One of a new collaboration.
I'll share more once the project is far enough along that discussing it will not frighten it back into hiding.
The Field Note - Very Cool Technology from a Company I Don't Trust
My day job continues to grow more intense as I move deeper into AI work.
Outside of work, I've started experimenting with Meta's new personal AI agent, Muse.
Wow.
It has some impressive abilities.
Muse is more than a chatbot that answers questions. It can operate a browser, fill out forms, work on projects after you close the application, and connect to services such as email and calendars. It can remember goals, suggest actions, and ask for approval before doing sensitive things such as sending an email or making a purchase.
That is where AI becomes considerably more useful.
It is also where trust becomes considerably more important.
According to Meta, Muse operates inside a dedicated virtual machine with a separate security agent watching what it does. Users control which services it can access, can review an audit trail, and can revoke permissions.
Those sound like sensible safeguards.
I still don't trust Meta.
Not almost.
Not a little.
I want to trust it because Muse is extremely cool. Unfortunately, Meta's history has not earned that trust, and connecting a personal AI agent to my email, calendar, browser, and financial accounts requires considerably more than a corporate promise that everything will be fine.
This may become one of the defining problems with personal AI agents. The most capable assistant will eventually need access to the information and systems we are least comfortable sharing.
An AI that cannot access anything can only provide advice.
An AI that can access everything might actually get the job done.
The important question is not merely whether an AI agent is intelligent enough to help us. It is whether the company operating it has earned the right to receive the keys.
Capability and trust are two entirely different engineering problems.
Meta appears to have made considerable progress on the first one.
It has a lot of work to do on the second.
Read Meta's announcement about Muse.
Murphy Was an Optimist
The AI will do exactly what you told it to do using the account you forgot it could access.
Behind the Wire
For the moment, the collaboration is moving to the top of the writing stack.
Steeds of Courage, Book Three of Marines, Luck and Oorah, remains with the publisher after adjudication and should be the next book released.
I also have no shortage of older novels and short stories waiting for attention. AI is helping me catalog the archaeological site that used to be my hard drive, but organization has an unfortunate side effect.
I can now see exactly how many unfinished projects are judging me.
Some will have to wait.
The Deep North will not conquer itself.
Reader Fire Mission
What would an AI agent have to prove before you trusted it with your email, calendar, browser, or credit card?
Would strong technical safeguards be enough?
Would the company behind it matter more?
Or are there some things you would never allow an AI to touch?
Hit reply and let me know. I'm genuinely curious where people draw that line.
Thanks for reading-and thanks for being here.
Until next Tuesday, keep your powder dry and your batteries charged.
-Bill
0 comments