Same here - 12y. I've learned so much and one day hope to contribute back something significant to the community but haven't found a footing just yet.

You don’t have to contribute something significant. I don’t think it’s about how much one contributes, or now important it is. If you have an opinion - give it. If you have a question - ask it. If you have a criticism - tell it (respectfully). I often get the most out of HN when I’m asking questions. The best part about this place is that people answer them. Or gives you the background context that wasn’t in an article.

This post is a good start!

Upstate South Carolina as well

I've been doing it since 1998 in my bedroom with a dual T1 (and on to real DCs later). While I've had some outages for sure it makes me feel better I am not that divergent in uptime in the long run vs big clouds.

Are you still on a dual T1? that's gotta be expensive

(and on to real DCs later) would imply their bare metal is now located in a data center.

really should stop skimming the comment when i find a part to comment on <facepalm>

Not OP, but I use Codex for back-end, scripting, and SQL. Claude Code for most front-end. I have found that when one faces a challenge, the other often can punch through and solve the problem. I even have them work together (moving thoughts and markdown plans back and fourth) and that works wonders.

My progression: Cursor in '24, Roo code mid '25, Claude Code in Q2 '25, Codex CLI in Q3 `25.

Cursor for me until 3-4 weeks ago, now Codex CLI most of the time.

These tools change all the time, very quickly. Important to stay open to change though.

For bare metal I’ve been using tier.net to get 192 GB RAM, 4TB NVME and 32 cores for $219/mo.

Data centers all over the country and I get to locate under 10ms from my regional audience.

Just a data point if you want some bigger iron than a VM.

Note this affects TETRA which is not used in North America. Most US systems use P25 which is not mentioned in the article.

Northern California services use P25 but with encryption turned off. They also have analogue repeaters. Presumably because that way they can still use old radios and don't have to worry about key rotation.

The audio quality on the analogue signal is a lot better than the P25 version, which is often harder to understand.

The P25 audio is digital, so it degrades worse than analog audio when signal quality is bad. You more quickly get to a point where it's not understandable at all, particularly if it's the signal coming from a handheld radio that's some distance away from the receiver.

The analog repeater is likely getting a high quality feed from the digital system, and then that's being broadcast at decent power from a tall antenna. It's starting with the best audio the system has to offer, has advantages in how it's broadcast, and degrades more gracefully.

I capture a lot of the P25 traffic in my area and there are times I can understand the dispatcher side of a conversation fine but the other end is too far away / too weak a signal and is unintelligible. The dispatcher's signal is coming from a fixed tower transmitting at higher power than a portable radio can manage.

Not like there’s not enough problems with P25… until the day they can deploy LLE (link-layer encryption) across all P25 systems, there will always be a way to gather some kind of intelligence about the system and its radio traffic.

(And the fact that it’s taking so long to implement link layer authorization, barely a scratch in the security dent…)

Thank you! This just saved me after closing laptop and losing a chat in VS Code. Cool feature and always a place where Clause Code UX was behind chat - being able to see history. "/continue" saved me ~15 minutes of re-establishing the planning for a new feature.

Also loving the shift + tab (twice) to enter plan mode. Just adding here in case it helps anyone else.

Claude code should really document this stuff in some kind of tutorial. There’s too much about code that I need to learn from random messages on the internet.

I too have noticed variability and it's impossible to know for sure but late one Friday or Saturday night (PST) it seemed to be brilliant, several iterations in a row. Some of my best output has been in very short windows.

I feel this too. I swear some of the coding Claude Code does on weekends is superior to the weekdays. It just has these eureka moments every now and then.

Claude has been particularly bad since they released 4.0. The push to remove 3.7 from Windsurf hasn’t helped either. Pretty evident they’re trying to force people to pay for Claude Code…

Trusting these LLM providers today is as risky as trusting Facebook as a platform, when they were pushing their “opensocial” stuff

I just had some issue with RLS/schema/postgres stuff. Gemini 2.5 Pro swung and missed, and talked a lot with little code, Claude Sonnet 4 solved. O1 Pro Solved. It's definitely random which of these models can solve various problems with the same prompt.

> definitely random which of these models can solve various problems with the same prompt.

Yeah, this is borderline my feeling too. Kicking off Codex with the same prompt but four times sometimes leads to for very different but confident solutions. Same when using the chat interfaces, although it seems like Sonnet 3.7 with thinking and o1 Pro Mode is a lot more consistent than any Gemini model I've tried.