bob1029

22,381 karmajoined il y a 9 ans

Open to new opportunities.

You may contact me using my email address, encoded as b64:

cm9iQHplbmlmeS5pbw==

Submissions

Pseudo-Chinese

en.wikipedia.org

5 pointsby bob1029il y a 3 mois0 comments

IBM 4769

en.wikipedia.org

2 pointsby bob1029il y a 4 mois0 comments

Evolution "Doesn't Need" Mutation. Here's Proof. - Blaise Agüera y Arcas [video]

1 pointsby bob1029il y a 5 mois2 comments

Recursive Knowledge Synthesis for Multi-LLM Systems

2 pointsby bob1029il y a 5 mois0 comments

WebView2, a browser for WinForms in .NET 5

grantwinney.com

2 pointsby bob1029il y a 5 mois0 comments

One Killed in train derailment near Barcelona days after deadly rail collision

1 pointsby bob1029il y a 6 mois0 comments

How to build agents with filesystems and bash

1 pointsby bob1029il y a 6 mois0 comments

Recursive Language Models W: Alex Zhang [video]

1 pointsby bob1029il y a 6 mois0 comments

Union Pacific 4014

en.wikipedia.org

5 pointsby bob1029il y a 6 mois2 comments

Making Deltron 3030 [video]

1 pointsby bob1029il y a 7 mois0 comments

Can a Modified Rivian Take the EV Cannonball Record? [video]

2 pointsby bob1029il y a 7 mois0 comments

Bepuphysics 2.4 [video]

1 pointsby bob1029il y a 8 mois0 comments

EA has released the Godot-based Battlefield 6 Portal SDK

5 pointsby bob1029il y a 9 mois0 comments

Effects of the entropy source on Monte Carlo simulations

2 pointsby bob1029il y a 10 mois0 comments

"Code Your Own Engine": Gearbox CEO Responds to Borderlands 4 Criticism

2 pointsby bob1029il y a 10 mois0 comments

comments

The natural gas is already being flared off to protect the environment (CH4 is a much worse greenhouse gas than CO2).

The key with this project is that instead of wasting the thermal energy from this unused gas, we feed it into power plants that are constructed nearby, eliminating the need for new pipeline infrastructure and providing some actual value.

You could go on to argue the oil fields themselves should cease production. This is the only way to actually stop the methane coming out of the ground. Even then it wouldn't happen overnight.

I am seeing some bugginess in testing:

  Parameter: reasoning_effort
  
  Function tools with reasoning_effort are not supported for gpt-5.6-sol in /v1/chat/completions. 
  To use function tools, use /v1/responses or set reasoning_effort to 'none'.'

Official OAI .NET library. Even when I override the currently experimental [?] flag to 'none', it will still occasionally throw this error (about 5% of the time).

I hope we aren't trying to push customers off the chat completion endpoint... Responses endpoint looks great on paper, but the business wants more visibility and control over the reasoning process than this product currently offers.

Edit: This is broken in my VS copilot setup too.

https://news.ycombinator.com/item?id=48630029#48631807

This looks like mindless data tourism to me. I don't see much distinction between this and something like Spotify's annual "wrapped" feature. The information theoretical equivalent of junk food.

avant-hierdiscuss

Infrasound from ventilation systems can be rough. Too much static pressure can create a lot of weird effects. I've reduced the ECM fan speed in my furnace to mitigate most of the noise. The system was replaced in 2020 and the installers didn't bother to change any of the factory defaults. I can understand why this happens too. The dip switches are absolutely tiny. I don't know if I could manage to change their settings after hauling the entire furnace into the crawlspace.

I'd estimate a majority of the R410a ducted systems have this kind of deficit in my area. The furnaces all ship from the factory configured by default for the max speed and very few installers actually bother with tools like manometers and static pressure readings. Napkin math says they had my 3 ton system blowing air like it was a 5-6 ton system.

avant-hierdiscuss

> Software of this kind of complexity has all kinds of behavior that has never been broken

This space of things is astronomically larger than the space of things expressly covered by any test suite.

"Program testing can be used to show the presence of bugs, but never to show their absence." -Edsger W. Dijkstra

avant-hierdiscuss

Project Kilby is probably the most intelligent approach to this problem so far.

https://www.chevron.com/newsroom/2026/q2/chevron-signs-20-ye...

The idea is to bring the data centers, power generators and energy supply together in the ~same physical space so the only thing you have to transmit is data. Moving energy is way more expensive than moving information.

il y a 3 joursdiscuss

I've been running a custom enterprise agent on 5.4 and it's been very good so far. I am looking forward to trying it with the monster model to see if we can approach some additional business cases.

I think if you are not seeing reasonable performance in your agent loops as of 5.5, it's likely there is a deficit with how the loop, prompt or tools interact with the environment.

il y a 3 joursdiscuss

The eCVT is definitely an end game transmission technology. Arguably the simplest drive train you can buy today short of pure electric. Even a manual transmission is more likely to fail.

il y a 3 joursdiscuss

It's pretty amazing what you can get in terms of power density these days.

https://www.amazon.com/Digital-Power-Amplifier-IRS2092S-Chan...

il y a 3 joursdiscuss

Here's the official docs if it helps

https://learn.microsoft.com/en-us/dotnet/csharp/roslyn-sdk/t...

il y a 3 joursdiscuss

I think semi-automation with contextual and domain-specific tooling is the key to the best quality outcomes.

For example, with browser automation, giving the LLM raw access to the literal DOM generally results in disaster for tasks that need to be stable across more than 5-10 interactions. The better approach is to write an intermediate layer that understands each view and can provide a list of tools that are precisely tailored for each case. E.g.:

  https://myapp/Login
  - <raw dom - hundreds of kb>
  - Available Tools: <arbitrary javascript>

vs

  https://myapp/login
  - We detected that this is the application's login page. 
  - It has the following visible elements:
    + Username
    + Password
    + Login Button
  - Available Tools:
    + PerformLogin
    + Quit

The later case takes a lot more effort, but it also reduces a Turing complete problem space into a binary decision at this particular step.

il y a 4 joursdiscuss

I don't think kernel goes far enough. We are pushing down the wrong path. We don't need kernel level anti cheats.

I think there should be real world legal consequences for exploiting information systems like this. "Gaming" shouldn't magically exempt a business sector from reasonable protection against fraud and abuse. Just because it's easy to do or involves bits in a computer (that isn't physically on the game developer's premises) should not magically exempt bad actors from prosecution. Yes, it's your computer, those are technically your bits, but this is about how these things are used and the context within which they are.

We have the CFAA in the US, but you'd need a fairly loose interpretation to cover most forms of cheating today. I don't really see the distinction between client and server when the actions taken on the client side cause a dramatic loss in quality of service for other clients connected to that same server or p2p relay.

South Korea has laws on the books as of 2016.

https://dotesports.com/overwatch/news/ow-hacker-sentenced-pr...

il y a 4 joursdiscuss

Honeypot & statistics are a much more interesting avenue of anti-cheat to me. The most egregious cheaters tend to get banned the most quickly with these techniques (e.g. a 200-1 K/D ratio = a lot of samples per unit time).

When it comes to cheating in video games, I subscribe more to the Dune philosophy than the LOTR philosophy. The ends (experience) definitely justify the means. Making sure that all cheaters are punished ~equally regardless of the effect of their infractions is fantastic in principle, but I'd rather if we could simply remove the ones that are obviously & openly ruining the experience.

False positives suck, but there is no reason to pander to the .01% performers if you are trying to keep a multiplayer game community alive. Esports has demonstrated this is a terrible approach. We had community servers that would ban people simply for being too good before all the matchmaking stuff started up. The anti-cheat was vibes based and arguably superior to anything going on today. It wasn't "fair", but it was generally a much better experience as a player.

il y a 4 joursdiscuss

> We will reduce management layers to no more than 5, and where possible, 3.

This reads like something from The Office.

il y a 5 joursdiscuss

> Computer science is neither about computers nor is it a science.

https://youtu.be/-J_xL4IGhJA

il y a 5 joursdiscuss

I pitched this to management as a way to get a very expensive fine tuning job done for free. They're not quite across that line yet but we're getting closer every day.

il y a 5 joursdiscuss

I'm struggling with the assertion that these models cannot provide reasonably deterministic guarantees.

I am using gpt to populate JSON objects conforming to a list of natural language constraints for purposes of generating fake customers. I am finding that gpt5+ never fucks up. Not even a little bit. I've ran this test hundreds of times with 20+ constraints and it's been perfect every time.

Stable information yields stable control flow. Humans are much more likely to forget one of the many constraints during testing. This happy mistake may incidentally cover an edge but it also means we lose coverage elsewhere.

I think whether or not the LLM should be allowed to directly author deterministic control flow (code) is mostly the same thing. If you have a lot of constraints you want to satisfy all at the same time, this can give you a hit very close to the ideal target very quickly. Not knowing exactly what you want is when the LLM takes you for a ride.

il y a 5 joursdiscuss

> I shot down similar arguments in favor of outsourcing overseas for years. Outsourcing any critical logic to an LLM is even worse.

Outsourcing to another continent of humans and supplementing workflows with LLMs are entirely different operational universes. I think it is fair to put them on the same spectrum, but they're really far apart.

I'd argue outsourcing is a far more aggressive abdication of ownership of the technology than bringing an LLM agent in house and having it light a few fires under a few asses.

il y a 5 joursdiscuss

https://en.wikipedia.org/wiki/Traffic_bollard