1. Some people want to have a very clear set of rules/process for which emails are "viewable" by an AI Agent, so a seperate "inbox" makes it a little easier for people to understand what an Agent is looking at (and add more customizability, controls, traceability, etc.)
2. Human/AI collaboration still requires some UX, we are not (yet) at the point of a fully automated agent that just does stuff and tells you what it did, I expect most apps need an AI-native version that have collaboration UX specifically made for giving feedback to an AI, approving what its done, etc. and MCP is just a way to specificy API access in a sense and not enough.
3. This first version is locally run and manually triggered (reactive) for fetching new emails, but we if you want to automatically run agents as new email comes in (proactive - which is our goal) we need a server to be processing new email and triggering the right agent, so this is one step towards that rather then just trying to give an MCP server to Claude or another assistant that is reactive instead of proactive.
- hosts + WireGuard AllowedIPs: This gives you a network-level tunnel. All applications on your system are forced through the VPN for that IP. You lose the per-app control.
- hosts + SSH Tunnel: This still doesn't fix the Host header problem. My remote reverse proxy gets a request for Host: localhost and returns a 404.
prxy solves both issues. It's an application-level proxy that correctly rewrites the Host header on the fly. It gives you the granular control that other methods lack. It’s a solution for a very specific problem, and other options are perfectly valid if you don't have the same constraints I did of course :)
I did not forget. The data is encrypted and keys are pinned on each side ahead of time. If the concern is hiding the source IP then one could route through multiple L4 vips using HAProxy on nodes in other countries. With time and money they may be able to penetrate each layer but one has to be worth the time and money.
This rings true to myself. While I thought I changed for the better, I did, but sacrificed all the things I really enjoyed doing. Best thing I did after 15 years was find myself again.
What I can’t discount is I’m a better version of myself than I was prior.
To be brutally honest, the problem sounds like a you problem.
You have a business, but no way to receive a sms. You had an empty account, you lost it? You used an email account you hardly check. Amazon isn’t this massive company because they are closing accounts with servers, that’s google cloud.
Now you spent a whole $30 when your ready again open a new account.
Your an edge case, and not doing any favors to make your life easier.
I may be wrong, but I think you are focusing too much on the practical consequences in this one case.
These consequences indeed are minimal - but that was due to luck, really, where I had not used the account.
If I had, and I had servers running, I would have been what on the face of it appears to be blackmailed into handing over sensitive, personal data.
There is also the question of AWS's procedures. One day of notice to close an account is not reasonable. It is a concern for any open accounts - on the face of it, it implies the need to check email from AWS every day and be ready to hand over any information AWS ask for, or your servers will shut down the next day.
What happens when you get sick? have covid, as I did? pow - servers all gone!
There's no discussion possible here, either, as Support take a few days to reply.
I may be wrong, but this is a concern.
I think you have to admit there is the possibility here for people to lose their accounts, without any chance of discussion, because they are unwell - and that is crazy; it's a flawed procedure.
"I was hit by a bus, in hospital with two broken legs and concussion, but my first priority was checking my AWS email every day! thank God AWS at least give one day of notice!"
(And the strangeness of two emails in the same day, with differing notice periods. What was going on, behind the scenes?)
I'm not looking to try again to open a new account, because I expect this to happen again.
>”Don’t fucking simp for Amazon. People do things different ways (hell, Amazon themselves barely have a phone number), and absent some indication of fraud their behavior is unreasonable.”
Easy solution for you. dont like their policies? dont do business with them.
if people create idle accounts on my own service id delete too. good news is you dont have to use it. :)
Yes, but there was a step change in the market between when he signed the deal and now. The price looked generous when he signed (which is not irrelevant; it was an offer the board essentially couldn't refuse, once his financing came in), but it looks absurdly high now. He likely expected it to be flat-ish or down a bit as the deal progressed toward closure, but not nearly this far down.
Oh could you give it a try now? I just woke up and got around to fixing some of the scaling issues - it's not perfect but at least it's only failing 5% of queries when I load-tested with locust. Will get around to better fixes later!
something along those lines, this gives everyone the ability to be involved and have a chance of winning the project and the poster to get what they pay for.
