WhyIsItAlwaysHN

"Social engineering is all you need"

That's an interesting interpretation of open source.

The article states that the particular item is a clear sign of fraud. If that was true, then it should be treated in a special manner. A more paranoid bank could enforce it without adhering to this guidance of multi-factor detection.

It isn't though, so balancing it with other rules is fine.

Exactly, hopefully this is not an autoblock in the future.

Tolls, revenue taxes, ever stricter rules that cause tickets despite technology getting better.

My experience was with building a new computer. I had gotten all of the parts needed to completely upgrade my pc, keeping only the case.After installing all of them, the computer refused to boot and just showed a generic error with a blinking light. I tried multiple things to troubleshoot like taking out each ram stick, different hdd, reseating cables, different hdd, change to internal gpu, nothing worked.

Then I fell asleep and when I woke up in the morning, I had a very strong feeling that the issue was definitely the memory and I had to move the ram to the set of slots for the other channel. And that was the issue, broken slots.

What's impressive is not that it worked, but that after waking up I had no doubt that was the problem, even though practically I couldn't conciously identify why I was so confident about it.

Maybe this is an unpopular opinion, but this seems to just create the search string in the url, aka youtube search already supports these features.

If people were really looking for exact title search they could write "term".

It is definitely true that youtube's search is optimized for engagement, but going through a separate ui just to search it seems a bit redundant, especially if after I click search I have ti deal with youtube's UI.

This result sounds very unsurprising at this point of having models that can reliably use tools.

Some part of RL training must focus on the length of responses. I would also guess that Anthropic and OpenAI have an incentive to optimize response length without sacrificing user satisfaction/retention.

For example, I would be more satisfied if claude code didn't execute a side-effect free script that produces no output. Embodying the concept of silence is semantically close to predicting the output of an empty program, so it's more efficient to say nothing.

Even in the past though similar tests gave output like says nothing. I think that points more towards optimizing for less tokens than the implied special understanding by the latest models.

Location: Copenhagen, Denmark (UTC+1) Remote: Yes (Europe timezones) Willing to relocate: No Role: UX Researcher Experience: 3+ years Email: rvbceilx0@mozmail.com

I research how people actually use AI products — then make them better. Cognitive science background (MSc), 3+ years on B2B/SaaS at a major tech company. Led the discovery research that took an AI agent from "what if" to production. Ran seven evaluative studies for gen-AI features now serving 100K+ monthly users. Also did a stint as a business analyst in financial services and led an AI data annotation team at a health-tech startup.

Methods: interviews, usability/concept testing, surveys, JTBD, journey mapping, quant in R/Python. Thesis on cognitive biases in LLMs. Looking for a team that uses research to make decisions, not to validate them.

Location: Copenhagen, Denmark (UTC+1) Remote: Yes (Europe timezones) Willing to relocate: No

Role: UX Researcher (mixed-methods / product discovery)

Experience: 3+ years

Email: rvbceilx0@mozmail.com

UX researcher with a background in cognition and communication, 3+ years on data- and AI-driven B2B/ERP products. I run mixed-methods studies (interviews, usability tests, concept tests, surveys, diary studies, JTBD, journey mapping, basic telemetry) and turn messy workflows and stakeholder questions into clear, actionable recommendations for product, design, and engineering. I’m mainly looking for interesting, challenging work with teams that are willing to listen to research; I’m flexible on domain and stage.

The OP just writes well. Also an llm is unlikely to write "thru"

I hope you're hiring for an agent developer position, otherwise it's not going to be a very effective signal

Location: Copenhagen, Denmark (UTC+1) Remote: Yes (Europe timezones)

Willing to relocate: No

Role: UX Researcher (mixed-methods / product discovery)

Experience: 3+ years

Email: rvbceilx0@mozmail.com

UX researcher with a background in cognition and communication, 3+ years on data- and AI-driven B2B/ERP products. I run mixed-methods studies (interviews, usability tests, concept tests, surveys, diary studies, JTBD, journey mapping, basic telemetry) and turn messy workflows and stakeholder questions into clear, actionable recommendations for product, design, and engineering. I’m mainly looking for interesting, challenging work with teams that are willing to listen to research; I’m flexible on domain and stage.

My own plug, translate between SQL dialects, state stored in URL so you can share it:

https://sqlscope.netlify.app/

Location: Copenhagen, Denmark (UTC+1)

Remote: Yes (Europe timezones)

Willing to relocate: No

Role: UX Researcher (mixed-methods / product discovery)

Experience: 3+ years

Email: rvbceilx0@mozmail.com

UX researcher with a background in cognition and communication, 3+ years on data- and AI-driven B2B/ERP products. I run mixed-methods studies (interviews, usability tests, concept tests, surveys, diary studies, JTBD, journey mapping, basic telemetry) and turn messy workflows and stakeholder questions into clear, actionable recommendations for product, design, and engineering. I’m mainly looking for interesting, challenging work with teams that are willing to listen to research; I’m flexible on domain and stage.

Ideally it's not weeding out but distributing into education paths which fit every student.

From my experience studying electrical and computer engineering, I definitely prefer that they chose to put hard electrical engineering courses in the first semesters because I knew immediately not to focus on them because I didn't like them.

They will not choose to write it. Would you work on something consistently if nobody cared about it?

There needs to be a reward for doing essays. That reward can be emotional eg. "the teacher I respect liked my essay" or "my essay was read in class" or "the teacher gives feedback that makes me feel a sense of growth". In that case, maybe kids will do it.

However, I think it's hard for a teacher to inspire respect to a classroom and the difficulty scales with the number of people in the class, so grades are used as a hack.

Outages would not be picked up

You might want to add a simple description of each company. I don't know half of them and in this dopamine shot optimized format, I don't really have the patience to research every one of them in a different tab.

Some groups will be more disadvantaged than others by being investigated. For example for welfare, I expect fraudsters to have more money to support themselves or less people to support (unless the criteria for welfare is something unexpected). So I'd say that there also needs to be more protections than just providing money.

Nevertheless the idea of giving money is still good imo, because it also incentivizes the fraud detection becoming more efficient, since mistakes now cost more. Unfortunately I have a feeling people might game that to get more money by triggering false investigations.

Or like a go beginner, which is fine

What I was saying is that you wouldn't use a raw LLM (so 506 tokens to get an answer). You would use it with web search so you can get the links.

The LLM has to read the websites to answer you so that significantly increases the token count, since it has to include them in its input.

But that was my point, then you need to include the entire websites in the context and it won't be 506 tokens per question. It will be thousands

There's something I don't get in this analysis.

The queries for the LLM which were used to estimate costs don't make a lot of sense for LLMs.

You would not ask an LLM to tell you the baggage size for a flight because there might be a rule added a week ago that changes this or the LLM might hallucinate the numbers.

You would ask an LLM with web search included so it can find sources and ground the answer. This applies to any question where you need factual data, otherwise it's like asking a random stranger on the street about things that can cost money.Then the token size balloons because the LLM needs to add entire websites to its context.

If you are not looking for a grounded answer, you might be doing something more creative, like writing a text. In that case, you might be iterating on the text where the entire discussion is sent multiple times as context so you can get the answer. There might be caching/batching etc but still the tokens required grow very fast.

In summary, I think the token estimates are likely quite off. But not to be all critical, I think it was a very informative post and in the end without real world consumption data, it's hard to estimate these things.

Good point, and much harder to challenge. If the majority is against an authoritarian there's protests and sabotage of social structures. If the majority oppresses a fringe group, it's often socially encouraged

What would you need to tackle this assignment?

I think you're right to complain about the design of the question. It used Google's disadvantage in that it couldn't process visual information to add difficulty to the task.

However I'm pretty sure there's many of these challenges in an average stem subject. An example would be information dense diagrams which describe some chemical process.

I'm genuinely curious to understand how someone manages to understand these without sight. On my side, I can barely visualize so it was always extremely hard to decipher them and even worse remember them.

Is the probability of lost data zero across eg. millions of documents?

I see there's a 30% redundancy per document, but I'm not sure every frame in a h265 file is guaranteed to have more than 70% of a qr code being readable. And if it's not readable, then that could mean losing an entire chunk of data.

I'd definitely calculate the probability of losing data if storing text with a lossy compression.

Is this more efficient than putting all of that in say a 7z archive?

I'd expect video frames to be maximally efficient if you sorted the chunks by image similarity somehow.

Also isn't there a risk of losing data by doing this since for example h.265 is lossy?

I love the idea of the colored links for navigation in your summary. Thanks for the inspiration!

He is the CEO of a company, he can personally ask someone to do it for him.