So I just asked Codex desktop to connect it for me-- and it turns out the latest desktop app has the hotkeys/bindings defined for the Micro already. In a couple of hours I had the core options bound and over the weekend I got to full LED status control.

The code is Windows stuff and probably not the hardware you have on hand, but that almost doesn't matter anymore, right? I'm more sharing as inspiration for people to hack their own devices. Once we see someone else do it, surely connecting to your random ESP32 touch-screen device or some XBox controller will offer you some convenience or novelty.

The biggest things I love for my personal workflow: 1. LED status of my running tasks and clicking them to go straight to the chat when it has a question or is done-- makes the multi-tasking so much easier. 2. Flipping a switch to start dictation and flipping it off when done. I dunno why this is so satisfying, but it is. 3. Having a dial to change reasoning level quickly (in my own harness I moved to multiple buttons since I change it so often), but the dial makes Codex desktop swapping so much more convenient. 4. Dial-select quick responses (e.g. "Make it so.") for all the stuff and common skills I use.

The code itself may not be super useful, but the project might give you or your agent some ideas on ways to connect these things to your own device or secondary machine.

Let me know if you have questions, I just wanted to share in case someone else found it handy.

Comments URL: https://news.ycombinator.com/item?id=48997577

Points: 2

# Comments: 0

Regarding the rest, it seems like a reasonable approach at first tinker.

Similar fun concept as the cataclysm library for Python: https://github.com/Mattie/cataclysm

As most of us here can see, for many tasks now you don't really need to worry that you have the exact right syntax. I think you still need expert precision when it matters immensely, but we all develop tools, scripts, layers and the like that manual precision isn't necessary.

I used a local Meross install on my old garage doors, time to break them out, but ugh...

I'm definitely curious on the context window increase-- I'm having a hard time telling if it's 'real' vs a fast specially trained summarization prework step. That being said, it's been doing a rather solid job not losing info in that context window in my minor anecdotal use cases.

I'll have to try that when I get back to my desk!

To be fair, that's what pretty much every person does. The bar does seem pretty high if we need more than that (especially if not specifically trained on a topic). It's not a universally perfect expert servant, but I've been exploring the code generation of GPT4 in detail (i.e. via the 'cataclysm' module I just posted about). In 1 minute it can write functions as good as the average developer intern most of the time.

We're keeping score in a weird way if we're responding quickly with it needing to "code without subtle but important errors". Because that's the majority of human developers, too. I've been writing code for 30 years, and if you put a gun to my head, I would still have subtle but important flaws in every first typing of any complex generated code.

I'm not saying you're bashing it, by the way, I get your point, but I do worry a bit when the first response is citing that the SOTA models get things wrong in 0-shot situations without full context. That's describing all of us.

   from cataclysm import doom

   # App gets the img file from the command line and saves it as a new file at half size with _half appended to the name
   doom.resize_app()

Any function you give it, it generates and executes. No need to think about what you need, it'll make it for you.

I haven't taught it to use itself yet if a module isn't installed, but that's next on the agenda once GPT4 isn't so slow.

Video intro here: https://www.youtube.com/watch?v=ZK8fUuQDgZ4

Some random examples when using global cataclysm:

  >>> uhoh = first_prime_with_3_digits()
  >>> print(uhoh)
  101

  >>> scary_words = get_italicized_phrases_wikipedia(page="Global catastrophic risk", only_lowercase_words=True, min_length=6)
  >>> print(scary_words[:5]) 
  ['existential risks', 'existential catastrophe[16]', 'global', 'terminal', 'permanent,']

You'll need a GPT4 API key-- it works with ChatGPT 3.5 turbo but that code is a little less reliable (and it can't generate as lengthy code).

It has two modes-- the ridiculous 'global' cataclysm mode for use in notebooks/interpreter that lets you hand-wave any function name and it'll cook it up for you. This is a bit scary (a typo turns into random code execution), so the second mode is just to use 'doom.' in front of any function name and it'll generate/cache its best guess for the right code given where it's called in the call stack (and its name/types/parameter names/etc).

A bit extreme, but a small concept to help us adjust to the impending cataclysm ahead of us. :)

Comments URL: https://news.ycombinator.com/item?id=35388270

Points: 7

# Comments: 3

Why do you think that? Competition? Can you elaborate?

Pretty sure there will be a thousand great libraries for this soon.

If high-end security firms and fairly diligent government agencies were infiltrated, why would we think that smaller dev toolchain organisations not founded as security organisations will somehow be less likely to be targeted and become a vector for introducing supply chain attacks. Sophisticated attackers will go after any soft underbelly or pore they can find, and there's no reason not to believe they'd put significant effort into quietly abusing Jetbrains security just like they did with Solarwinds. I'm less worried about the "Russian" red scare mention other than it may give non-US organizations a few more opportunities to inject badness that we can't get visibility on.

Bottom-line is that it would be the holy grail and are we treating it as the high priority target it is? Having worked for dev tool companies in the past, I know they are a lot more worried about innovation than about their own internal processes.

For all sorts of objectives, government spy agencies have made a science of repeatedly putting spies, moles, and compromised individuals into places of power within other major governments (and even other spy agencies). Why wouldn't the same tactics be used in much easier circumstances like placing (or 'compromising') agents who are engineers, administrators, recruiters, or contractors at important tech companies?

Hiring: Do any of you know of major tech/infrastructure companies that consider such things while recruiting/contracting?

Threat modeling: Is anyone here including such possibilities in their security threat models for workflows, networks, and software? Rogue employees seeking personal gain (and sometimes corporate espionage) has come up in threat models we've done for years, but we never really considered imaginary infiltrators from one's own government, nor more complicated scenarios (such as an entire contracting company being a front for a 3 letter agency). The typical NDAs, contracts, and legal defenses clearly don't offer much in the way of protection/demotivation for those cases.

This might be too paranoid of a topic, but I'm genuinely curious if people are seeing signs that tech companies are mulling over such possibilities now (especially given how much heat companies are taking for government involvement).

Comments URL: https://news.ycombinator.com/item?id=7002376

Points: 2

# Comments: 0