A2W

Try Innkeeper 2FA:

[Denislav Gavrilov @kuberdenis]

Innkeeper v1.1 (now on the App Store):

Connectors are now supported:
- OAuth support: google, github, ms365, gitlab, discord
- PAT: jira, slack, notion, linear, stripe, cloudflare, vercel

Try Innkeeper: https://t.co/s5OhWZyk71

[Denislav Gavrilov @kuberdenis]

10th place. (right behind Expo Go 🫪)

Some features I’m working on:
- oauth connectors
- sticky session
- cloudkit <> tailscale toggle
- contextual profiling 2fa (will take time)
- desktop app secrets management (debating)

Try innkeeper! Give your agents Auth with minimal config, but more importantly - securely!

https://t.co/3d8n7q2WHU

[Denislav Gavrilov @kuberdenis]

Demo for Innkeeper, the first 2FA for agents

try it at https://t.co/s5OhWZyk71

/i get a food delivery call midway thru presenting just ignore that/

[Denislav Gavrilov @kuberdenis]

Well, now it is No. 18 !

[Denislav Gavrilov @kuberdenis]

Innkeeper landed in charts!!

Feels good. Stressful, too lol

[A2W @a2wdotio]

https://t.co/MJH5At3QLz solves this!

- local, on-device store
- LAN transport (works with tailscale)
- auth responses are never passed
- 2fa for auth/orized actions

it is also free as in completely free :)

[Chaofan Shou @shoucccc]

I bought a Fable dataset from one of the top Chinese LLM routers yesterday.

With just 6TB data, I can take over 7 Chinese/CIS gov entities &amp; 19 top Chinese firms like Xiaomi, Huawei, NIO, Minimax using SSH keys, VPN configs, Aliyun keys, GitLab tokens sent to the router.

[A2W @a2wdotio]

Testing out innkeeper. In the field

[Denislav Gavrilov @kuberdenis]

Running the UBI AGI test.. feels strange

Will report

[Denislav Gavrilov @kuberdenis]

Getting there 🫡 follow @a2wdotio

[Denislav Gavrilov @kuberdenis]

MFA for agents. Releasing soon

https://t.co/Z2uD95gsfE

[Denislav Gavrilov @kuberdenis]

2FA for agents

follow @a2wdotio we are just getting started

[A2W @a2wdotio]

[6/6] This is not evidence about Anthropic’s margins. List price is revenue forgone, not what it costs them to serve, and at $3,977 a month nobody runs an agent like this at all /Denis here, they do.../.

The whole post, charts and stuff:
https://t.co/lqTVX5fxhg

[A2W @a2wdotio]

We ran an ~always-on autonomous agent for 20 days on @Anthropic’s Max 20x with one question in mind:

“How much does Anthropic lose on a fully-utilized Max 20x”

Here’s what we found: [1/6]

[A2W @a2wdotio]

[5/6] 94 readings across two weeks, one line through the origin: $18.36 of list-price work per meter point.

A full week at 100% is ~$1,836, so ~$8,000 a month against a $200 bill. That is the 40x.

Depends on your model mix though. All-Fable that week is ~$1,260, all-Opus ~$4,490.

[A2W @a2wdotio]

(5/5) Try Rouse at https://t.co/6g0as2AJSg

[A2W @a2wdotio]

Introducing: https://t.co/6g0as2AJSg

Rouse is a memory framework that we believe takes a good step towards making multi-session model "awareness" &amp; sub-agent orchestration better.

[A2W @a2wdotio]

(4/5) And here is a technical writeup on how Rouse works and why we are calling it a "memory framework"

https://t.co/yEHNgPBJra