I tried gemma-4-26B-A4B just to see if it could help me read/sort my emails on a relatively under-powered setup (16GB VRAM + 32GB RAM) and it's not going well.. the model burns 24K tokens just on searching for the right tool and then dumps the email contents into context - i tried to get it to use code-mode to save context but the code-mode implementation can't save files so it was useless and im going to try to switch to "ssh-mode" into my devbox container. Still relatively new to this, so I'm probably doing something wrong
So there was a problem with gemma 4 when it comes to tool calling that Google apparently fixed like 2 or 3 days ago. I remember reading something about this.
They would have no internally/externally defensible justification to stop the launch as they are partners/part-owners of Anthropic. They would have to let the rank-and-file keep moving on the Fable launch.
IIRC Anthropic models haven't all been available on day 1, so it does feel like a deliberate choice, especially since they are partners/part-owners like you say. No one would have bat an eye about some corporate PR thing "blah blah mythos is too powerful and too intelligent and so we've decided to focus our capacity on Opus 4.6/4.7/4.8 for now until we have the proper safeguards in place blah blah"
This is corporate Game of Thrones, nothing more. Amazon, maybe in alliance/deals with others as well saw an opportunity to hurt their rival. Or maybe they were instructed to report this by the WH themselves. Hegseth and the WH will happily take any excuse to hurt Anthropic after the confrontation with DOW, being the vindictive cronies they are.
If you don't trust them, then no policy is enough. Technically everything you send to the model could be stored by them. Personally I do worry about that especially as an average consumer not an enterprise, no one is looking out for us and we don't get any guarantees. But enterprises will get the right treatment because they would find out and sue Anthropic if they lied.
I made a little tool where I can just drop HTML into a folder and it will deploy it either to my internal Caddy or publicly on Cloudflare based in folder. Can be a single html or a folder.
CF pages still required too much confusing clicking around on a webpage for me. This way I can just point any little report or app at a directory and done.
There’s others that are more server shaped and tightly coupled - a pipeline to pull in all my data like Garmin, Twitter bookmarks, messages into a Postgres DB. Kind of a personal data warehouse that i can use with apps/automations, like alerting me if my sleep schedule is drifting, and a custom web interface to my Garmin data
I was trying WSL years ago and this is one of the reasons I just moved to a full linux server instead. We still have way too many problems interfacing across filesystems. I hope with AI we will see an iteration on ExFAT that has all the journalling, versioning etc. magic of modern FS' and can be adopted across all 3 OSes. Probably a long shot but I can dream :)
Yes that would be great. Right now, there are many applications that use pinned certificates to communicate to servers meaning there is literally no way to see the data your own device is sending/receiving from the internet. It's an insane thing that should be banned.
There is one way, you can modify the app or the OS to change which certificate is pinned, ignore all certificate failures, lie about the certificate in use, log encryption keys, or not even ask the app whether it likes the certificate.
I tried the VPS briefly, it didn't really solve anything for me. The personal assistant agent is only as useful as the data & tools it has, that's where the real risk is. Separate box gives you isolated FS but docker also does that very easily.
Docker is not a security boundary. It never has been, but given recent demonstrations of container escapes its even less of one than it ever was. If you want to properly contain a process it needs to be running in a VM of its own, or you need to accept that there's a risk of it escaping and ending up with more access than you planned.
Yes literally it threw an invoice from AWS into spam recently. Regularly throws financial docs into spam. Like you, i have a problem with false positives not false negatives.
I brought it up with support once and the domain was failing DMARC. Headers from the AWS email that has a spam score of 9.1! I don’t know much about email but something must be wrong there on Fastmails side.
reply