0o_MrPatrick_o0

Dvorak's curmudgeonly approach to tech journalism was deeply enjoyable. He was utterly un-sycophantic and a playful troublemaker. Condolences to his wife, family and fellow podcasters.

Thank you for sharing this observation.

I had a post on here this week that was briefly popular. There were numerous folks who posted that the material “wasn’t new.”

I felt like it should be implicit that people who already know what I was writing about are not the target audience. But here they were, commenting.

At the same time, the page got upvoted quite a lot and the comments were filled with folks who had lots of interesting reactions and additions. Despite the fact that this was “old news”, it seemed implicit that for many it was new news.

Sometimes we should trigger conversation. I believe we shouldn’t index on novelty- we should index on impact. Your post is a nice defense of those who discover and share what they discover.

I had an order of operations error. I better edit it because you’re the second person to get nerd sniped. Sorry friend- you are right

Reading these comments is so harrowing.

You are correct in my intentions on this post generally.

I want to highlight:

I want to measure performance of the LLMs over time- which includes assessing the quality of their outputs. I don’t perceive the reasoning output to be anything other than a measurable signal of possible drift in model performance.

Except it isn’t, because I’m only getting a low value summary of the thinking.

It’s like asking your buddy how fast he thought that last pitch was when radar guns are behind the plate.

Yeah, it’s a description related to what happened, but it’s not the thing I want to measure.

Awesome share! Thank you!

I missed my coffee! Ty! Five points to Slytherin.

My bad! 10 points for House Slytherin!

I want to measure performance drift over time.

Having access to the reasoning text and output would help with performance measurement.

It’d be nice to see some example code. Good writeup otherwise!

"Carlini had never before found a bug in Linux, or in Ghost. Now he had discovered many."

New guy learns nessus, now tells everyone at the bar he's basically Mr. Robot.

A pox on the labs and the government. InfosecDrama.exe just took out a frontier model because a noob learned how to use a tool.

Hey! Cool idea! I was lamenting the state of advertisements in biblehub and biblegateway and thought it’d be nice to see something like what you’ve done.

Where are you pulling your datasets from? Is there a Strong’s dataset available for integration?

‘The United States’ foreign router ban didn’t make a whole lot of sense, and today may not change that.`

???

No obvious reason? What if the Executive Branch is a dog chasing cars?

It’s just doing things.

If you need evidence of Amazon’s ability to wield AI-driven development towards creative and productive ends…

Look elsewhere.

I think it’s unwise because Model reliability is transient.

When the models have an off day, the workflows you’ve grown to depend upon fail. When you’re completely dependent on Anthropic for not only execution but troubleshooting- you’re doomed. You lose a whole day troubleshooting model performance variability when you should have just logged off and waited. These are very cognitively disruptive days.

Build in multi-model support- so your agents can modify routing if an observer discovers variability.

I’ve been building my own version of this. It’s a bit shocking to see parallel ideation.

FWIW- IMO, being locked into a single model provider is a deal breaker.

This solution will distract a lot of folks and doom-lock them into Anthropic. That’ll probably be fine for small offices, but it is suicidal to get hooked into Anthropic’s way of doing things for anything complex. IME, you want to be able to compare different models and you end up managing them to your style. It’s a bit like cooking- where you may have greater affinity for certain flavors. You make selection tradeoffs on when to use a frontier model on design & planning vs something self hosted for simpler operations tasks.

It may be possible that this has changed. Last year I built a device and crashed into the bridging weirdness when I wanted to use upstream dhcp. There was/is some funkiness lurking with bridging wifi to Ethernet- in particular with broadcasts that traverse the bridge.

> spectfully the scenario you want to present seems to change.

Man that is totally a fair point.

I feel like I’ve struggled with the tutorials on these configs so many times in my life that I’ve kind of munged several ideas together here. There’s so much subtlety to the iptables/nftables rules that I failed to understand for so long, that I forgot that some folks might not understand that WiFi has specific weirdness. You’re right- I open with routing as a topic, but I’m in a very specific nuance right away.

Respectfully- I don’t think this statement applies to the scenario I presented.

“The Linux box instantly turns into a router as soon as you run `sysctl net.ipv4.ip_forward=1`, because the default policy for FORWARD table is ACCEPT.”

In the setup I presented, we are bridging an Ethernet and a WiFi network. This would be desirable if you wanted to use an upstream dhcp server for your WiFi clients- or if you wanted to avoid double nat’ing.

In 802.11 infrastructure mode, a station can only send frames with its own MAC address. The AP won’t accept or forward frames from unknown MACs. So you can’t transparently bridge Ethernet devices’ MAC addresses through a WiFi client interface. This is why we need hostapd.

In every other circumstance- I think your statement holds.

I tried to do some weird alerting on new MAC addresses and ran into this weirdness. Bridging WiFi and Ethernet gets weird.

I wouldn’t call it a cpe unless it translates the connection from the clec/co into IP.

IMO, it’s a plain old router/switch/bridge.

Customer premises equipment.

People who use this term are in telco.

Thank you for the pointers!!

Some notes on using SSH-AGENT to protect your private keys from AI Agents while allowing them ssh access using those keys.

because you're trusting Claude's implementation of hooks, which may be disastrous if they have a defect.

Amazing tip! Thank you!

Hey man- sorry for the lack of recognition. Timing is a bit of luck. Good writeup!

w/r/t .git being mounted read write- yeah, there's risk here. It's a tradeoff. I want my agents to be able to commit code- which means they need to be able to write to the dir.

Thanks for the --new-session parameter suggestion. Great add!

Heya- The reason I add this directory is because Claude needs read/write permissions for getting new auth tokens.

Without this, you'll have to re-login to Claude every time. Breaks the speed of development.

I'm going to do some experimenting to see if I can make this bind more precise.

Using IaC tools gives you the following advantages:

1. IaC gives you Idempotent solutions- which is advantageous over an agent. What if the agent crashes half way through deployment procedures? How will you reliably resume an interrupted install?

2. IaC gives you reproducible builds

3. IaC gives you ability to install tools in a way that can be tested for compliance with any deployment standards you have

W/r/t frontier models:

Just tell them to go install stuff. They already have so much in their training corpus that you literally do not need to create this.

7b-14b parameter self hosted models may get some benefit from your approach. I find self hosted is less reliable for tasteful approaches. Micromanagement yields better results.

Author should explore Ansible/Puppet/Chef.

I’m not sure this solution is needed with frontier models.

You’ve got my intent correct!

Where I’m at with #2 is the agent builds a prototype with its own private session credentials.

I have orchestration created that can replicate the prototyping session.

From there I can keep final build keys secret from the agent.

My build loop is meant to build an experiment first, and then an enduring build based on what it figures out.

Great question! You might enjoy this writeup, which in one section explores avoiding the use of shell variables that are not exported as a method of mitigating this risk.

https://linus.schreibt.jetzt/posts/shell-secrets.html