joshribakoff

Thats how you end up goodhearts lawing your way into simplicity eg duplicate code. Thats like the whole takeaway of the article at least for me.

“raise 17m to try to kill off and extract value from popular open source tool”

Pound foolish and folly

After the January triple billing it took me 20 emails to get a human who alleged he didn’t see the charges. After that, i received only AI responses again. Upon opening credit card disputes they finally replied to say they cant help now because of the disputes. I won the disputes.

The TUI makes more sense to humans who don’t understand the difference between a human and a machine.

> I'm not going to dump tens of GB of stuff to a log file

In the same vein as the parent comment, the curiosity is why you would vibe code a solution instead of reaching for grep.

> We exclusively use 1M internally, so we're dogfooding it all day

That is so out of touch. Customers do not exclusively use 1M. This is like a fronted developer shipping tons of unused Mb and being oblivious because they are on fast internet themselves.

They don’t enforce or even default to 2fa to change the account email. In addition, they have no process to get a human to reverse account takeovers. Just a web form that tells you to call a number that redirects you back to a web form

I cant even read this because most of the text is outside my phones viewport. Please test your stuff before posting it here.

After a while many people then realize this often produces worse results by injecting additional noise in context like the overhead of invoking the gh cli and parsing json comments or worse the mcp.

But they get the dopamine loop of keeping the loop alive, flashing colors, high score/token use, and plausible looking outputs — so its easy to deceive oneself into thinking something remarkable was discovered

The “secret sauce”? RPI (research plan implement) is far from a secret concept. Heck, its in the official docs! Not secret.

“never forgetting”? Its still a probabilistic model.

You’re framing it like you discovered some secret techniques that eludes 100s of millions of users and overcame the architectural limits of the model.

SFs de facto stance (in practice due to the dynamic between law enforcement agencies) seems to be that generally an individual has right to relieve themselves even on a sidewalk (even pretextually) and being visible does not in itself arise to a crime like indecent exposure, without additional factors (such as the stereotypical “trench coat flashing” without pretext).

Thats the way SFPD worded it to me after repeatedly failing to act on my reports (such as a woman urinating in front of my child on a sidewalk near an sf muni bus stop). Nudity is legal. People have a right to bodily function. Another factor is response time, for this priority the response time can be over 12 hours (if they respond), and obviously they’re not gonna skip trace someone for urinating on a sidewalk after they left the scene.

Most event organizers like bay to breakers race or outside lands provide facilities to mitigate burdens on the community. The city provides its own portable facilities in hotspots. Tesla could easily do it, but is notorious for ignoring regulations or just doesn’t care.

Bad code has real world consequences. Its not limited to having to rewrite it. The cost might also include sanctions, lost users, attrition, and other negative consequences you don’t just measure in dev hours

Its a large leap from “we made a config driven diagram tool and trained an llm on that config” to “all apps will disappear”. If you’re predicting such grand claims please be more precise than “AI” which is a term we cant define.

While nothing fundamentally changes i have found an increased need for tests and their taxonomies — because the LLM can “hack” the tests. So, having more robust tests with more ways to organize and run the tests. For example instead of 200 tests maybe i have 1,200, along with some lightweight tools to run tests in different parts of the test taxonomies.

A more concrete example is maybe you have tests that show you put a highlight on the active item tests that show you don’t put the highlight on the inactive items, but with an llm you might also want to have tests that wait a while and verify the highlight is not flickering on and off overtime (something so absurd you wouldn’t even test for it before AI).

The value of these test is in catching areas of the code where things are drifting towards nonsense because humans aren’t reviewing as thoroughly. I don’t think that you can realistically have 100% data coverage and prevent every single bug and not review the code. It’s just that I found that slightly more tests are warranted if you do want to step back.

It does claim the US went to great lengths to dismiss the victims for a decade, while being in possession of the device. That raises the question of what incentives the US would have to deny its existence. To me, that was the story.

Earlier this year a recruiter contacted me about a staff role there. Within minutes of the call they asked if I’d consider senior instead. I decided not to proceed.

This has been widely known in bodybuilding and powerlifting circles, people abusing performance enhancing drugs eat things like oats to mitigate the harmful effects of the drugs on their cholesterol, and regularly do blood work to monitor it and see that it is working.

Very cool idea. If your product is about video, please fix your video players. I cannot even seek on my touch screen.

I agree with all of that. Also consider that there is an argument that the guard rail only stops the good guy. Not saying that’s a valid argument though.

You wouldn’t say that rolling dice is dangerous. You would say that the human who decides to take an action, depending on the value of the dice is the danger. I don’t think AI is dangerous. I think people are dangerous.

Dario’s opinion on safety won’t necessarily matter if he’s not even in the room. This move keeps him in the room.

Cool. I did this exact same thing at Cruise to prove and defend my approach to monitoring e2e safety/latency in the remote assistance tool.

In particular, i established that sudden changes in connection speed create latency in the video channels that do not correlate with the data channels

All that an agent has to do now is write one line of code to log it at the top of your program.

An LLM is just a compressed version of the web. In this context, I don’t see a meaningful distinction between “distill” vs “compress”.

Very cool. I have a similar side project for scraping youtube playlists and aggregating open source texts. Mainly materials for computer science, system design, and DSA (data structures and algorithms).

On GH as joshribakoff/leetdeeper

There are examples of labs banning these use cases for sure, as well as the presence of terms and conditions allowing them to ban you for merely “competing” with them. If you’re building, it could be worth locking in a contract first.

Very cool. The biggest questions someone skimming would likely be why the letters are in this order, and how this is consumed (eg ios app?). You may answer those details but they were not front and center to me.

> Some observers have pointed out cases where CCC [sic] appears to regenerate artifacts strongly resembling existing implementations, including standard headers

Its not just that, we just had another thread here on HN recently discussing how the LLMs reproduce the entire work of Harry Potter with ~99% accuracy when prompted to do so with a jailbreak. This seems to contradict the “it’s remarkable progress” statement from the top of the article.

> a framework for applying code changes across hundreds or thousands of repositories at once

Statements like this raise fair questions. Is there code duplication across 1,000s of repos, and why respond by increasing surface area further with bespoke tooling?

Check out primeageons 99 prompts. The idea, as i understand it, is you scope an agent to implementing a single function at a time with firm guardrails. So something in between yolo agents and rudimentary tab complete