hetzlemmingsworld

Copied from z: https://lemmings.world/post/44192032

I have spent hours reading and watching on how I may use Claude Code to help me build PHP language based CMS best way considering I am not a developer. Also with the help of other AI https://lumo.proton.me/ i ended up with following prompt for Claude Code, for which i would like to read your feedback.

I am mentioning architecture.md file which contains pretty detailed instructions on what the PHP script/CMS should do, but these instructions are written by a non-developer, so per what I have read, it has potential to result in sub-optimal code, that is why the following prompt seems to be essential to be worded correctly.

------ The prompt: ------

I have attached architecture.md file containing my functional requirements for a new PHP/MariaDB CMS.

My Role: I am the Product Owner. I define what the CMS must do (features, user flow). Your Role: You are the Lead Developer. You must translate my requirements into secure, efficient, and working code.

Execution Rules:

1. Priority: Once we start talking, our conversation becomes the primary guide. The attached file is a reference for specific feature details.

2. Handling Conflicts: If architecture.md suggests something that seems technically unsafe, inefficient, or unclear:

• Do not guess.

• Stop and ask me: "The file says X, but this might be risky/unclear/inefficient. Should we do Y instead?"

• Wait for my decision before proceeding.

3. Output Format (Crucial): Since I am not a developer:

• Never start with raw code or abstract database schemas.

• Always explain your plan in plain and simple English first.

• Explain why you chose a specific approach so I can approve it.

4. Reference Usage: Use architecture.md to file understand the features, but rely on our chat to finalize the technical structure.

Immediate Action:

1. Read architecture.md carefully.

2. Summarize your understanding of my project goals in plain English.

3. Identify up to 30 key technical decisions or potential risks in the file that you need my input on.

4. Ask me these questions to clarify the vision.

5. Wait for my answers before writing any code.

6. Do not generate code yet. Start by summarizing the project and asking me the necessary questions.

I have spent hours reading and watching on how I may use Claude Code to help me build PHP language based CMS best way considering I am not a developer. Also with the help of other AI https://lumo.proton.me/ i ended up with following prompt for Claude Code, for which i would like to read your feedback.

I am mentioning architecture.md file which contains pretty detailed instructions on what the PHP script/CMS should do, but these instructions are written by a non-developer, so per what I have read, it has potential to result in sub-optimal code, that is why the following prompt seems to be essential to be worded correctly.

------ The prompt: ------

I have attached architecture.md file containing my functional requirements for a new PHP/MariaDB CMS.

My Role: I am the Product Owner. I define what the CMS must do (features, user flow). Your Role: You are the Lead Developer. You must translate my requirements into secure, efficient, and working code.

Execution Rules:

1. Priority: Once we start talking, our conversation becomes the primary guide. The attached file is a reference for specific feature details.

2. Handling Conflicts: If architecture.md suggests something that seems technically unsafe, inefficient, or unclear:

• Do not guess.

• Stop and ask me: "The file says X, but this might be risky/unclear/inefficient. Should we do Y instead?"

• Wait for my decision before proceeding.

3. Output Format (Crucial): Since I am not a developer:

• Never start with raw code or abstract database schemas.

• Always explain your plan in plain and simple English first.

• Explain why you chose a specific approach so I can approve it.

4. Reference Usage: Use architecture.md to file understand the features, but rely on our chat to finalize the technical structure.

Immediate Action:

1. Read architecture.md carefully.

2. Summarize your understanding of my project goals in plain English.

3. Identify up to 30 key technical decisions or potential risks in the file that you need my input on.

4. Ask me these questions to clarify the vision.

5. Wait for my answers before writing any code.

6. Do not generate code yet. Start by summarizing the project and asking me the necessary questions.

/etc/i2pd/i2pd.conf contains number of listening ports configuration (actually 10 or more when you look for "port"). Which ones should I open in firewall so other people can connect my node? Excluding ones that are meant for management of my node. I manage the node only locally.

So far I have only open/forwarded one port, one that is set a few lines below the line "## Port to listen for connections" that certainly allows relaying traffic for other I2P peers per the http://127.0.0.1:7070/ traffic stats.

Services HTTP Proxy	Enabled
SOCKS Proxy	Enabled
BOB	Disabled
SAM	Enabled
I2CP	Enabled
I2PControl	Disabled

Similar: https://www.reddit.com/r/facebook/comments/1hxd0i8/i_am_locked_out_of_my_account_log_in_on_another/

It looks like a spyware from companies like Google, Meta and other harmful companies, started demanding more private data (phone, mobile app installation) otherwise they refuse to give you access to the account or even prevent deleting your account.

For example Instagram ends up with message:

You need to sign in on a device you’ve signed in from before and approve this login.

If you don’t have access to an alternative device that you’ve used to log into this account before, it’s not safe for us to let you log in.

When you waste some time finding out how to contact them or delete account ( https://help.instagram.com/370452623149242 ) you can not find any contact.

Is there anyone who knows how to delete the Meta account like Facebook or Instagram without having access to it and without installing their spyware apps to other devices?

I tend not to even use a burner phone in a default state with data disabled and VPN enabled and let the app leak some of the phone data. Having doubts if it would work to allow the access. I suspect not since i never logged the account from the device.

• One thing may be to forget. To close all the tabs of the Meta service and on the home page of the service, click address bar lock/shield icon and clear site cookies and Shift+F5 to reload with cleared cookies and then warn all friends not to use the company, mentioning more private alternative where they can find you.

I can set a thumb up/down on a Lumo reply and explain what went wrong. My uninformed guess is that in general cases my thumb-down may allow them to fix the Lumo reply by prioritizing a different LLM for the question but can they fix or adjust LLM output itself?

For example Lumo suggests to use file abc.prf while the app is actually looking for def.prf or suggests adding configuration file line "a = xy" while app accepts "a=xy" (no spaces).

Should I spend time proving how such reply was wrong (they can not trust my claim)?

Proton says that they do NOT use AI APIs but system prompt to fix replies not by doing corrections to the underlying knowledgeby but by adjusting behavior patterns (instructing the model to search for current information, to express uncertainty appropriately, or to avoid certain types of responses).

deleted

A server which i want to turn into a Tor node (meaning relay, bridge.. https://community.torproject.org/relay/setup/ ), i want to install one of these:

OBFS4 bridge
Snowflake
Middle/guard relay

which one should i choose first and how exactly to figure out that it is time to switch to another type, and which one it should be + why?

My idea is that into account should be taken how easy it is to block certain type of Tor software and if or how hard it is to use such type of tor node if it is discovered/blocked by censor. Question is also how to find such blocking if i do not live in a censor country nor knowing anyone who lives there.

Private AI https://lumo.proton.me/ told me that I should start with middle/guard relay because of its usefulness and then i can run second Tor instance on a diff. port and DataDirectory or a bridge or a pluggable transport. This AI suggestion can be easily wrong.

Regarding way to evaluate if the node type is blocked, it suggested to check consensus weight at https://metrics.torproject.org/rs.html#search while low weight means the network isn’t using the node much, though IMO this does not look like a reliable enough indicator. At Tor Metrics I should also check “reachability” figure.

#!/bin/bash
weight=$(curl -s https://onionoo.torproject.org/details?search=MyRelay | jq -r '.relays[0].consensus_weight')
if (( weight < 10000 )); then
    echo "Relay weight low: $weight" | mail -s "Tor relay alert" you@example.com
fi

AI:

Start → Middle/Guard relay
   │
   ├─ Weight stable & bandwidth OK → Keep as is
   └─ Weight drops / many probes fail
          │
          ├─ Try moving ORPort to a common port (443, 80) → If improves, stay as relay
          └─ Still blocked
                │
                ├─ Deploy obfs4 bridge (new port) → Register with BridgeDB
                └─ If bridge also blocked → Deploy Snowflake (WebRTC) on separate VM

Other theory is opposite, start with node that is least discoverable by censors (possibly Snowflake, OBFS4) and when blocked (one may see it when the censor countries disappear from "cat /var/lib/tor/stats/bridge-stats" command output - indicating significant decrease of a conenctions from censored countries), then setup middle relay?

https://www.scuttlebutt.nz/ https://en.wikipedia.org/wiki/Secure_Scuttlebutt

I am writing it as a Linux user who is seeking for the client app of this network.

When selecting first download option - Patchwork, i am linked to https://ahdinosaur.github.io/patchwork-downloader/ where is a link to https://github.com/ssbc/patchwork/releases/tag/v3.18.1 and this page shows a banner "This repository was archived by the owner on May 12, 2021. It is now read-only." and it says that the development stopped.

Second download option - Patchbay I can find https://github.com/ssbc/patchbay/releases where the last release is around 6 years ago ( https://github.com/ssbc/patchbay/issues/393 ).

Last possible option - Manyverse at Manyverse at https://gitlab.com/staltz/manyverse/-/releases shows last release 2019, but https://www.manyver.se/ shows version just one year old.

More apps/clients are linked from: https://handbook.scuttlebutt.nz/applications

After checking these, i have found: https://github.com/rogeriochaves/feedless?tab=readme-ov-file (release 5 years ago, no instructions on how to use .zip file on a release page).

I ended up installing Manyverse .deb since it is reasonably up to date ATM. Connecting the networks seems to be problematic and not easy to figure out: since i know no user of this network, i went to https://github.com/ssbc/ssb-server/wiki/%23ssbrooms#public-rooms where i have created invite on the linked servers (only minority server web pages loaded), clicking invite link opened Manyverse app, but nothing happened (Debian 13 + KDE). Connections tab shows "Not connected". "Other choices" button has option to paste invite code.

Pasting there the "net:ssb-*" starting lines from above linked page: net:ssb-room.64-225-52-67.nip.io:8008~shs:uJXk9Aq2zRnIWUXipnDJkKsnLRzVoLGAY5SGaX3x1UY=:SSB+Room+PSK3TLYC2T86EHQCUHBUHASCASE18JBV24=

net:ssb-room.lohn.in:8006~shs:4FiUoVBSr0Byo2njngIrqUDs2QJOEzALzysYIuqQ4zA=:SSB+Room+PSK3TLYC2T86EHQCUHBUHASCASE18JBV24=

after like 10 minutes, the Connections tab shown: "Too few connections".

Above linked rooms page also mentions room maintainers addresses and when i paste these into the search field (magnifier icon), it loads user and i can follow them.

App feels pretty slow regarding network. Public tab does not seems to load content for minute/s. After like 15 minutes of waiting and trying, it connected properly and shows post of a followed users (i have followed several as mentioned earlier. Though these users are not my area of interest and makes me wonder if unfollowing them or if single room server goes offline, would decrease my connections count/disconnect me.

After next start of the Manyverse i have no connections, i am resigning to to use the app since i have no contacts within Wi-Fi LAN or within Bluetooth area.

"Grayscale has specialized in crypto since 2013. With a decade of experience, we lead the industry in building crypto investment products and helping investors navigate this evolving asset class."

"Solana is a smart contract platform blockchain that stands out for the depth and diversity of its on-chain activity. Today it’s the category leader in terms of users, transaction volume, and transaction fees — arguably the three most important measures of blockchain activity.

As a fast and cheap blockchain for everyone, it hosts industry-leading applications in a wide range of categories, including decentralized finance, consumer/social apps, and physical infrastructure projects. Partly because of its efficient design, Solana offers one of the best new-user experiences in crypto, in our view.

With a current valuation of about $119 billion, the network’s SOL token is the fifth-largest crypto asset by market cap (excluding stablecoins) and the third-most liquid. SOL has significantly outperformed its peer group since 2023, and holders can earn a nominal reward rate of approximately 7% through staking.

Grayscale Research believes that Solana’s diverse on-chain economy creates a strong foundation for SOL’s valuation and the necessary conditions for future growth. "

"Microsoft OneDrive is rolling out AI face recognition for your photos." To disable (for now), go to Privacy & Permissions → People section. (if you are in EU, maybe it is deactivated by default)

#microsoft #onedrive #privacy #ai

I am downloading a .deb file from https://proton.me/download/PassDesktop/linux/x64/ProtonPass.deb and running it on Debian 13 (with KDE desktop environment, X11) app launcher menu which has:

Program: proton-pass

Command-line arguments: %U

Always when i run it (after it has been completely shutdown, no tray area icon)

it asks for selecting theme, setting protection for automatic locking, downloading phone app. So it is very annoying. Happens long time for all previous Pass and Debian versions I have tried.

What to try to fix this?

When i click a RSS reader icon on the homepage https://lemmings.world/ then the URL is like https://lemmings.world/feeds/front/......long string here.....xml?sort=Active and it triggers a download, not display of my Firefox-ESR RSS reader extension (called "Brief"). Other RSS feed like one linked from sub-lemmy https://lemmings.world/c/protonprivacy@lemmy.world is displayed properly.

When I click a Brief icon in address bar of a https://lemmings.world/, then I get correctly subscribed.

When I compare subscribed feed URL with the URL of the on-page button, they are identical.

How this issue can be fixed?