HNHacker News
TopNewBestAskShowJobs

twosdai

347 karma · joined February 3, 2020

YC S22

Social stuff: https://www.linkedin.com/in/daniel-wasserlauf-b12681101/

Email: daniel.wasserlauf@gmail.com

submissionscomments
twosdai··on X Didn't Fix Grok's 'Undressing' Problem. It Just Makes People Pay for It
It, the difference between calling child pornographic content cp vs CSAM, is splitting hairs. Call it CSAM its the modern term. Don't try to create a divide on terminology due to an edge case on some legal code interpretations. It doesn't really help in my opinion and is not a worthwhile argument. I understand where you are coming from on a technicality. But the current definition does "fit" well enough. So why make it an issue. As an example consider the following theoretical case:

a lawyer and judge are discussing a case and using the terminology CSAM in the case and needs to argue between the legality or issue between the child being real or not. What help is it in this situation to use CP vs CSAM in that moment. I dont really think it changes things at all. In both cases the lawyer and judge would need to still clarify for everyone that "presumably" the person is not real. So an acronym change on this point to me is still not a great take. Its regressive, not progressive.

twosdai··on Cowork: Claude Code for the rest of your work
Didn't actually check out the app, but some aspects of application state are hard to serialize, some operations are not reversible by the application. EG: sending an email. It doesn't seem naively trivial to accomplish this, for all apps.

So maybe on some apps, but "all" is a difficult thing.

twosdai··on X Didn't Fix Grok's 'Undressing' Problem. It Just Makes People Pay for It
> no children were sexually assaulted

Generating pictures of a real child naked is assault. Imagine finding child photos of yourself online naked being passed around. Its extremely unpleasant and its assault.

If you're arguing that generating a "fake child" is somehow significantly different and that you want to split hairs over the CSAM/CP term in that specific case. Its not a great take to be honest, people understand CSAM, actually verifying if its a "real" child or not, is not really relevant.

twosdai··on TimeCapsuleLLM: LLM trained only on data from 1800-1875
> but it definitely has some bias.

to be frank though, I think this a better way than all people's thoughts all of the time.

I think the "crowd" of information makes the end output of an LLM worse rather than better. Specifically in our inability to know really what kind of Bias we're dealing with.

Currently to me it feels really muddy knowing how information is biased, beyond just the hallucination and factual incosistencies.

But as far as I can tell, "correctness of the content aside", sometimes frontier LLMs respond like freshman college students, other times they respond with the rigor of a mathematics PHD canidate, and sometimes like a marketing hit piece.

This dataset has a consistency which I think is actually a really useful feature. I agree that having many perspectives in the dataset is good, but as an end user being able to rely on some level of consistency with an AI model is something I really think is missing.

Maybe more succinctly I want frontier LLM's to have a known and specific response style and bias which I can rely on, because there already is a lot of noise.

twosdai··on Worst of breed software
As an outsider of Big co's. I always felt that if youre not on one of the 10-20 awesome product teams. Eg, Google maps, aws lambda, windows core os. Something along those lines. It seems like a territory for justification Olympics.

Just my view as a dev who's largest co was like 500 people. ~100 engineers.

twosdai··on Ask HN: Any Microsoft employees/devs here? What's happening to Microsoft?
If youre able to do some networking and tell people verbatim what you just wrote. I think that will help a lot more than grinding kernel logic or whatever.

If youre not considering it, I would reccomend trying to find a local hacker club like defcon or 2600. There are usually a few embedded folks who go or are affiliated with those.

twosdai··on Show HN: Foundertrace – chain of YC startups founded by its employees
Yc founder, there's not really any pressure like this that exists. Basically i think it's a proximity thing and consistent exposure to a specific section of the startup world that makes this happen.

Basically all those people would go all the same networking events sponsored by yc. There's not pressure so much as these people all have frequent and paid for, opportunities to "hang out" and talk about tech stuff together. Its possible you could define this as "pressure" but I think it's more of who you hangout with rather than some top down implicit force.

twosdai··on Strange.website
Marquee is the greatest html tag of all.
twosdai··on DHS Says DHS-Certified Real IDs Too Unreliable to Confirm U.S. Citizenship
There's gotta be some Midwest or southern state that absolutely doesn't care.
twosdai··on Show HN: OpenWorkers – Self-hosted Cloudflare workers in Rust
Completely agree. There are some apps that unfortunately need to care about some level of security isolation, but with an open workers they could just put those specific workers on their own isolated instance.
twosdai··on Humans May Be Able to Grow New Teeth Within Just 4 Years
I wonder if at the time the "world of tomorrow" is how we think of it now with fondness.
twosdai··on An 11-qubit atom processor in silicon with all fidelities from 99.10% to 99.99%
Photonic computing has a lot of practical applications for signal transfer already.

Basically anytime we send a signal across a large fiber optic cable we need to convert signal from light back to electricity and that requires some level of photonic computing. Its used at scale today. https://www.ebsco.com/research-starters/science/photonics

However I suspect that you mean photonic computing where a computer on chip device uses photons instead of electrons to communicate. In which case, as far as I know is still research phase.

twosdai··on Log level 'error' should mean that something needs to be fixed
I understand where the author is coming from but frankly I think this design pattern is not really correct.

Obviously this depends on teams, application context and code bases. But "knowing if action needs to be taken" can't be boiled into a simple log level for most cases.

There is a reason most alerting software like pagerduty is just a trigger interface and the logic for what constitutes the "error" is typically some data level query in something like datadog, sumologic, elastic search, or graphana, that either looks for specific string messages, error types, or a collection of metric conditions.

Cool if you want to consider that any error level log needs to be an actionable error but what quickly happens is that some error cases are auto retry able due to infrastructure conditions that the application has completely no knowledge of. And to run some sort of infrastructure query at error write time in code, eg

1. Error is thrown 2. Prior to logging guess/determine if the case can be retired through a few http calls. 3. Log either a warning or an error

Seems to be a complete waste when we could just write some sort of query in our log/metrics management platform of choice which takes into account the infrastructure conditions for us.

twosdai··on AWS CEO says replacing junior devs with AI is 'one of the dumbest ideas'
This is a wrong take. People learn plenty while using AI. it's how you use it that matters. Same issue happened years ago if you just copied stack overflow without understanding what you were doing.

Its no different now, just the level of effort required to get the code copy is lower.

Whenever I use AI I sit and read and understand every line before pushing. Its not hard. I learn more.

twosdai··on No AI* Here – A Response to Mozilla's Next Chapter
In the digital world, we should be able to go back from output to input unless the intention of the function is to "not do that". Like hashing.

Llms not being able to go from output back to input deterministically and for us to understand why is very important, most of our issues with llms stem from this issue. Its why mechanistic interpretabilty research is so hot right now.

The car analogy is not good because models are digital components and a car is a real world thing. They are not comparable.

twosdai··on VPN location claims don't match real traffic exits
A more general case is for legal and SLAs. If a company uses one of these vpns to make sure their traffic only travels through a specific legal path, and then it's found that their traffic entered a different territory, there can be a lot of consequences.

The case I can think of most accessible would be anything that streams copywriten video.

twosdai··on Show HN: Spotify Wrapped but for LeetCode
I think for some people, they treat it like the daily crossword. For seniors, I've known a few that would do problems in a different language so they can get some basic exposure to it. Eg. If you program node js all day, you'd do the daily problem in kotlin or rust.
twosdai··on The Junior Hiring Crisis
Writing automated tests is not the same thing as testing it. Senior engineers write a lot of code, write some tests, do some manual validation. But its usually far from complete.

Other roles in a software team can handle making a complete test suite for a project, and making sure that it covers all of the required end user functionality. In some shops they will push all of these together into a "fullstack" kind of role, in other software shops they will have dedicated team members who's job it is to handle some of this burden. Interns historically filled in the glue for this.

twosdai··on Paged Out
Super cool. Loved reading this.
twosdai··on The Junior Hiring Crisis
Writing unit tests, manual validation work, manual testing. Automating Deployments of infrastructure, DNS work, tracking down annoying one off bugs, fixing and validating dependency issues.

Basically this type of maintenance work for any sufficiently complex codebase. (Over 20k LOC)

When I was an QA intern / Software Dev Intern. I did all of that junk.

twosdai··on I don't care how well your "AI" works
I really agree with your point. I think that this forum being hackernews and all though lends itself to a slightly different kind of tech person. Who really values for themselves and their team, the art of getting stuck in with a deeply technical problem and being able to overcome it.
twosdai··on Ozempic does not slow Alzheimer's, study finds
This is anecdotal, I don't have proof but it's something I think is somewhat related. Is that obesity and nuerodegenerative diseases are somewhat related. So that's a guess as to why some people might have though a weight loss drug would potentially be related in some way with alzheimers.

This could just be false though, I can't recall where I heard this information. So do some searching before quoting me.

twosdai··on We Induced Smells With Ultrasound
Not joking, my mom has asked me every year since I became a "tech person" when I am going to make smellovision. This is perfect for her.
twosdai··on The New AI Consciousness Paper
It's magic. Chill out. It knows.
twosdai··on You can make PS2 games in JavaScript
Super cool. Thanks for sharing this. I've been looking for something like this for a while.
twosdai··on The US AI Bubble Reminds Me of the Eve of China's Real Estate Collapse
I think some of the valuation, or agreements made right now are based on future cash flow promises, which are assumed to be higher.
twosdai··on The politics of purely client-side apps
Build it and get your friends to join you.
twosdai··on Is it possible to have an LLM always give a consistent output?
Really appreciate this, this is exactly what I was looking for.
twosdai··on Is it possible to have an LLM always give a consistent output?
To clarify I am valuing consistency in the output to the token level, not the actual information of the content.
twosdai··on The end of the rip-off economy: consumers use LLMs against information asymmetry
There is a lot more astroturfing than you know. People with multiple accounts create question answer cases all the time to just talk about a product.
← PreviousPage 3 of 9Next →