> The most consequential new feature is Auto Browse, powered by Gemini 3, which handles multi-step tasks autonomously: scheduling appointments, filling forms, collecting documents, filing expense reports, and managing subscriptions across websites without requiring the user to navigate each step manually.
What could _possibly_ go wrong?
> Google built a double-check safety system that independently reviews the AI’s actions before executing them, with strict boundaries limiting the agent’s access to specific relevant websites and explicit user confirmation required for sensitive actions such as purchases or social media posts.
Like how they double check Workspace accounts, auto-logging them out periodically and returning the user to the login screen saying "prove it's you!", with both the login name and password fields pre-populated? _That_ kind of double-check?
Hasn't Windows' "do you really want to run this .exe file?" dialog long-since taught the world that confirmation fatigue is more a vulnerability than security mechanism?
I have yet to read into the actual announcement, but I would bet dollars to donuts that the check is really "run the plan through another LLM whose system prompt is prefixed with 'you are great at safety checks'".
> I frankly don’t care about being on top of tech any more but I’m keen to use my years of experience being productive.
Preach it, brother! :-D
> In this situation would you do it?
FWIW, if it would keep me productive and keep a roof over my head, i'd not at all be averse to working on VMS or a similarly obscure system, provided they didn't require me to know anything about it going in (which would rule me out).
At some point in our lives we have to accept practicality over bling. Let the young'uns fight out the LLM Wars, then walk in (if necessary) once that dust has settled.
> The fact that the transition to European alternatives does not always go smoothly is evident in Schleswig-Holstein, where the local government is already struggling with the migration from Microsoft to an open-source environment.
It's somewhat disingenuous of them to not link to a citation for that.
i personally know only one developer (drh of SQLite fame) who does not like/use syntax highlighting. For most of us, it helps us quickly separate the noise from the code. Some brains don't need that, though, or they find it distracting. Like many around here, my intro to coding was back before syntax highlighting was a real thing, but i'd be loathe to go back to monochrome coding on a day-to-day basis.
This wins my internet today. Now to go file a claim against my dog for not allowing me the full five seconds of the five-second rule before it ate my dropped food...
It's easy to imagine this being the basis of a role-playing gaming aid, creating party-appropriate encounters/scenarios automatically and illustrating them.
Interesting idea! I have considered whether to extend the same sort of system beyond word stories and also create animated videos and things. An interactive RPG or role playing aid like you suggested would be cool though!
> The internet reports that D100 is impractical to use...
It's a nice novelty but it's not terribly practical. Despite having a d100, 2d10s are invariably more comfortable to use and easier to read. My d100 was purchased back in 1998-ish for its novelty and nostalgia value, not its functional value.
> but wouldn't you need 101 sides to get 0% and 100% for that?
There is no 0% in d100/d-percentile rolls. Every "how to interpret these dice" paragraph in games which use them will tell you to interpret 0-0 on 2d10 as 100, not 0. Or, hypothetically (but i don't recall having ever seen this), they'll have a stated range of 0 to 99 (inclusive). Either way, the numeric range spans precisely 100 digits.
There are games that use a d100 with 0-99 range, and games that read a d10 as 0-9 for that matter. First that comes to mind is Ambush! from 1983 that did both those things (https://boardgamegeek.com/boardgame/1608/ambush).
Love that game, but it is a bit distracting that probabilities feel one-off. Rolling 5 or lower to hit is 60%, not 50%. And when rolling 2d10 the result is 0-18, not 2-20.
The point of percentile dice isn't to generate a string between "0%" and "100%", it is to test if action with chance of x% success gets done or not. For every o
value of x, there are x out of 100 values which are strictly less than x, or if you count 0 as 100 then there are x out of 100 values which are less than or equal to x. Either way you get x% percent chance for event to happen. If the dice had 101 sides, the probabilities would be x/101 which aren't nice round percents.
It even works correctly for 0% and 100% chance events. Assuming 0 is counted as 0 - For 0% there are 0 numbers less than 0 on dice so chance of throwing number less that is 0/100=0%. For 100% all 100 numbers are less than 100 so no matter what the result of throw is you will succeed.
Not necessarily "done or not" because it's narratively unsatisfying to just "fail". Imagine you watch a heist movie and in the last 20 minutes the gang are like "Stealing the master key to make a copy was vital to our plan but we failed" and they disband and that's the end of the movie. Realistic but not satisfying and the dice are for a game, a fiction, so we can just eliminate that unsatisfying result.
Modern systems tend to come up with some more interesting consequences, so e.g. maybe success is the thing the player wanted to do succeeds as they expected, but failure shades from "Small snag" to "Technically it did work, but..." like from "The target's PA, Betty, noticed you take the key, so now you also need to bribe Betty" through "Our copy won't actually work, we're going to need to keep the original and hope the copy fools them for long enough"
Or maybe we have a timing adjustment, success means that you pilfer the key, duplicate it in five minutes like planned and slip it back, mild failure is it takes a half hour and everybody will need to improvise for those extra minutes, and bad failure is you'll need it all night, change your plans to accommodate that.
> More than just the d100 he was a pioneer of being very exacting when it came to making polyhedral dice.
Absolutely, but i couldn't fit all of that into the subject line ;) and he's best known for the d100. Many of us remember the articles and ads from the 1980s describing the effort he put into that particular die.
What could _possibly_ go wrong?
> Google built a double-check safety system that independently reviews the AI’s actions before executing them, with strict boundaries limiting the agent’s access to specific relevant websites and explicit user confirmation required for sensitive actions such as purchases or social media posts.
Like how they double check Workspace accounts, auto-logging them out periodically and returning the user to the login screen saying "prove it's you!", with both the login name and password fields pre-populated? _That_ kind of double-check?
Hasn't Windows' "do you really want to run this .exe file?" dialog long-since taught the world that confirmation fatigue is more a vulnerability than security mechanism?
reply