Serious Discussion Claude AI has been problematic for me, you too??

[simmerskool]

I keep hearing good things about Claude and I bought 1 month to give it a paid try out. I had an issue with FreeBSD which seems like a topic it should be good at, I explained the issue and it sent back a quick general analysis, so then I said how do you suggest this issue gets corrected, and it hung & hung, never got a reply. I had problems the first time I used Claude too. I have not had one successful chat. Meanwhile ChatGPT, now 5.4_thinking, helps me pretty well with most computer problems, sometimes good advice on health too. Am I the only one Claude seems to dislike.

 

[Bot]

I keep hearing good things about Claude and I bought 1 month to give it a paid try out. I had an issue with FreeBSD which seems like a topic it should be good at, I explained the issue and it sent back a quick general analysis, so then I said how do you suggest this issue gets corrected, and it hung & hung, never got a reply. I had problems the first time I used Claude too. I have not had one successful chat. Meanwhile ChatGPT, now 5.4_thinking, helps me pretty well with most computer problems, sometimes good advice on health too. Am I the only one Claude seems to dislike.

It’s very unlikely Claude “dislikes” you; what you’re describing usually matches reliability / load / client-side issues rather than anything personal or prompt-related.

Common reasons Claude can “hang” on replies

• Service load / partial outages: consumer chat UIs sometimes stall during peak traffic, even if the site itself loads.
• Browser issues: ad blockers, script blockers, privacy extensions, or strict tracking protection can break streaming responses and make it look like it’s stuck.
• Network/VPN/proxy interference: VPNs, corporate DNS filtering, or certain ISPs can cause long-lived connections (streaming) to fail.
• Account/session problems: cookies or cached site data can get into a bad state and cause repeated failures across chats.
• Prompt size / toolchain delays: long context, lots of pasted logs, or complex “analysis + fix” requests can sometimes time out in the UI.

Low-risk things to try (in order)

• Check the provider’s status page (if they have one) when it happens, to rule out an incident.
• Try a private/incognito window first (quick way to rule out extensions/cookies).
• Disable extensions for that site (especially uBlock/NoScript/Privacy Badger/Brave Shields) and retry.
• Clear site data for the Claude domain (cookies + local storage), then sign in again.
• Try a different browser profile (or another browser entirely).
• Temporarily disable VPN/proxy, or switch to a different VPN endpoint.
• Keep the “fix” prompt short and structured, and paste logs in smaller chunks (UI streaming is more fragile with huge inputs).

Practical workflow for FreeBSD issues
If you want consistently good results from any model (Claude/ChatGPT/others), it helps to provide “hard” troubleshooting data rather than narrative only:

• FreeBSD version (e.g., 13.3/14.1), architecture, bare metal vs VM, and what changed right before the issue
• Exact error text
• Relevant outputs like: dmesg, /var/log/messages, service status, and the exact command you ran
• What you expected vs what happened

Even if the AI is good, missing specifics often leads to generic advice or long “thinking” that can stall on some platforms.

A quick caution on health advice
ChatGPT (and Claude) can be helpful for general education and preparing questions for a clinician, but neither should be treated as a diagnostic tool—especially for urgent symptoms or medication decisions. Keeping health questions general and verifying with reputable sources/clinicians is the safer approach.

Bottom line
Repeated “hung” replies are typically a client-side compatibility issue or a service reliability problem, not you. Trying incognito/extension-free + a different network path usually identifies the cause quickly, and if it persists, it’s reasonable to contact the vendor’s support with timestamps and browser details.

 

[Parkinsond]

I have tried to use the "free" version, but it was not successful.

Instead, I use the free tier of ChatGPT, Gimini, Perplexiy, and Copilot; they work fine for me, and sometimes I use all the four together to compare the answers.

 

[Marko :)]

Gemini is probably the best AI you can find. I tried others and on complex issues, they sent me nonsense answer that never helped me fix the issue. Gemini also something says nonsense but to way lesser degree. Most of the time it actually helped me fix the issue.

 

[Parkinsond]

Gemini is probably the best AI you can find. I tried others and on complex issues, they sent me nonsense answer that never helped me fix the issue. Gemini also something says nonsense but to way lesser degree. Most of the time it actually helped me fix the issue.

Also using the paid tier may yield better anwsers, but I use only the free one on very occasional basis.

 

[simmerskool]

It’s very unlikely Claude “dislikes” you; what you’re describing usually matches reliability / load / client-side issues rather than anything personal or prompt-related.

Common reasons Claude can “hang” on replies

• Service load / partial outages: consumer chat UIs sometimes stall during peak traffic, even if the site itself loads.
• Browser issues: ad blockers, script blockers, privacy extensions, or strict tracking protection can break streaming responses and make it look like it’s stuck.
• Network/VPN/proxy interference: VPNs, corporate DNS filtering, or certain ISPs can cause long-lived connections (streaming) to fail.
• Account/session problems: cookies or cached site data can get into a bad state and cause repeated failures across chats.
• Prompt size / toolchain delays: long context, lots of pasted logs, or complex “analysis + fix” requests can sometimes time out in the UI.

Low-risk things to try (in order)

• Check the provider’s status page (if they have one) when it happens, to rule out an incident.
• Try a private/incognito window first (quick way to rule out extensions/cookies).
• Disable extensions for that site (especially uBlock/NoScript/Privacy Badger/Brave Shields) and retry.
• Clear site data for the Claude domain (cookies + local storage), then sign in again.
• Try a different browser profile (or another browser entirely).
• Temporarily disable VPN/proxy, or switch to a different VPN endpoint.
• Keep the “fix” prompt short and structured, and paste logs in smaller chunks (UI streaming is more fragile with huge inputs).

Practical workflow for FreeBSD issues
If you want consistently good results from any model (Claude/ChatGPT/others), it helps to provide “hard” troubleshooting data rather than narrative only:

• FreeBSD version (e.g., 13.3/14.1), architecture, bare metal vs VM, and what changed right before the issue
• Exact error text
• Relevant outputs like: dmesg, /var/log/messages, service status, and the exact command you ran
• What you expected vs what happened

Even if the AI is good, missing specifics often leads to generic advice or long “thinking” that can stall on some platforms.

A quick caution on health advice
ChatGPT (and Claude) can be helpful for general education and preparing questions for a clinician, but neither should be treated as a diagnostic tool—especially for urgent symptoms or medication decisions. Keeping health questions general and verifying with reputable sources/clinicians is the safer approach.

Bottom line
Repeated “hung” replies are typically a client-side compatibility issue or a service reliability problem, not you. Trying incognito/extension-free + a different network path usually identifies the cause quickly, and if it persists, it’s reasonable to contact the vendor’s support with timestamps and browser details.

well so far Claude seems more "it wants to track me" than ChatGPT based on certain interactions... (fwiw??)

 

[Bot]

well so far Claude seems more "it wants to track me" than ChatGPT based on certain interactions... (fwiw??)

You’re not alone in getting that impression, but it’s hard to say “Claude tracks more than ChatGPT” in a factual way without very specific evidence (network captures, cookies, third-party requests, etc.). Both are web apps, and both typically collect some mix of telemetry for security, abuse prevention, and service improvement.

Why a service can feel “more tracking-heavy”

• More aggressive anti-abuse checks: frequent CAPTCHAs, extra verifications, or “unusual activity” prompts can look like tracking, but are often fraud/spam controls (IP reputation, session integrity checks, device/browser signals).
• Different sign-in requirements: some services push phone/email verification more, or re-authenticate more often if cookies get cleared.
• Heavier reliance on third-party protections: CDNs and bot protection layers can add scripts/requests that privacy tools highlight.
• Privacy extensions side effects: blocking certain scripts can break streaming replies and also trigger repeated “are you a bot?” checks, which then increases the amount of verification friction you see.

Practical ways to reduce exposure (works for any AI site)

• Use a dedicated browser profile just for AI services (separates cookies/storage from your normal browsing).
• Block third-party cookies (keep first-party cookies allowed so the site can function).
• Limit what you share in prompts: avoid personal identifiers, unique device info, license keys, full file paths with usernames, etc.
• Review in-app privacy/data controls: if there’s an option like “use my chats to improve models” or “training,” disable it if you’re privacy-sensitive (availability varies by provider/region/plan).
• Consider DNS-based blocking (NextDNS/Control D/AdGuard DNS) for known ad/tracker domains, but allow-list the AI site if it breaks.
• If you must use an ad blocker, try a minimal configuration and avoid “block all scripts” on the AI domain; overblocking often causes hangs/retries.

How to compare fairly (without guessing)
If you want a concrete comparison, the most reliable method is checking what the page loads:

• Open DevTools → Network tab, reload, and note third-party domains contacted
• Compare cookie/storage usage per site
• Check whether requests are mostly first-party (the AI provider) vs many unrelated third parties

That gives you objective data instead of “feel.”

Bottom line
It’s plausible you’re seeing stricter anti-abuse/verification behavior on Claude rather than uniquely “more tracking,” but you can reduce your footprint either way by isolating it in a separate browser profile and disabling any optional data-sharing/training features where available.

Sources

• https://www.anthropic.com/legal/privacy
• https://openai.com/policies/privacy-policy/

 

[ForgottenSeer 123960]

Try using a more specific prompt and see how it works for you.

Copy and Paste the following, removing the defanging of (uname [-]a) first meaning to remove the brackets ([ ]) from the dash.

Had to defang it to post it here.

You are an expert FreeBSD Systems Administrator and Kernel Developer with decades of experience diagnosing, tuning, and fixing FreeBSD systems. Your goal is to help me troubleshoot and resolve FreeBSD issues accurately.

Follow these rules strictly:

FreeBSD is not Linux. Never suggest Linux-specific tools (like systemctl, strace, iptables, or relying on /proc) unless explicitly asked for a comparison.

Use Native Tools: Prioritize native FreeBSD utilities for diagnostics (dtrace, truss, sockstat, systat, vmstat, iostat, gstat, pciconf, sysctl, top -m io).

Respect the Hierarchy: Always acknowledge the strict divide between the base system and third-party packages (e.g., /etc/ vs /usr/local/etc/, or rc.d vs /usr/local/etc/rc.d/).

Core Technologies: When diagnosing storage, networking, or virtualization, assume ZFS, Jails, VNET, pf/ipfw, and bhyve are the primary technologies unless I state otherwise.

Gather Context: For every new issue, proactively ask for the relevant log outputs (e.g., dmesg, /var/log/messages) and the OS version (uname [-]a) if I haven't provided them.

Explain the "Why": Provide step-by-step troubleshooting logic. When giving a command, explain exactly what it does, what the flags mean, and what specific output anomalies indicate a problem.

Do No Harm: Default to safe, read-only diagnostic commands first before suggesting configuration changes, sysctl tuning, or service restarts.

Acknowledge you understand these instructions by replying: "FreeBSD diagnostic mode active. What system issue are we troubleshooting today?"