themanmaran

Yea in one of the further Ender's Game novels they talk about this tech a lot for characters silently mumbling to the AI.

Do you have a way to demo the voice without downloading and running the model? The Huggingface playback doesn't work.

Yea even reading the first couple paragraphs I can't really get a sense of what it's trying to say. Feels like generic claude output and doesn't really communicate what the "crisis" is or how it's getting "worse".

Fun fact, there is a .gov effort to sell (or give away) a lot of old lighthouses in exchange for agreed maintenance. It's sort of a terrible deal, since you can't live in them or use them for commercial purposes, but you do still have to maintain it to a certain standard. But I'm sure someone is out there with the desire.

https://realestatesales.gov/our-listing/

> do you get laid off?

> Neither has anything to do with your skills at the actual job.

I hate to break it to you. But it definitely does...

Great to hear. A philosophy to keep in mind:

LLMs give back nothing but mountains of text. To provide value, you must provide something other than text.

Agreed! Clicked the link because I was excited. This is a real gap especially when you have like 10 different agent providers and no idea what they are doing. But the UI here is just bad.

IMO the correct UI for this like gather.town. Give me a top down pokemon view of all my agents running and a leaderboard for activity / most recently run.

I don't think me giving the model bad instructions is something a smarter model can solve. I use plan mode constantly (with opus/fable), and at least once a day I'll say something too vague or just dumb and it will sketch out the "wrong" solution in it's plan.

Which is fine because it just put together a plan and didn't spend 10 minutes rearchitecting everything.

I likewise don't understand this push. Surely it's a trivial feature to keep enabled. It mostly seems like advertising that "our models are so smart you don't need to plan". But I find it to be a great mental review process for my workflow.

It sounds like the Ant perspective is "you can just ask claude to plan", but at the same time that's a little more tedious than shift+tab

I have very fond memories of reading hitchhikers guide to the galaxy with a friend in highschool. The book was old and had already started falling apart, and so we just broke it up into chapters and would pass it to the next person one chapter at a time.

Monumint (YC W24) | https://monumint.com | Full Stack Engineers | SF In-Person | Full Time | $125k–$225k • 0.50%–1.50% Tech stack: TypeScript | Node | React/Next.js | Postgres | Docker | K8s | LLMs

We’re building the Voice AI infrastructure layer for financial services. Banks and fintechs use Monumint agent to automate loan origination, account opening, servicing and customer engagement.

You’ll work on:

1. Driving down voice latency to human-conversation levels while scaling to thousands of concurrent calls

2. Deploying agents that run durable workflows across days or weeks, maintaining state across calls, tools, documents, and human handoffs

3. Making AI agents reliable and observable in a financial workflows: tool calling, evals, guardrails, and deterministic fallbacks

We’re early, you’ll have a huge impact. Email me (my name at the URL) or apply here: https://www.workatastartup.com/companies/monumint

I don't think it's that binary. For the majority of websites, adding programatic access didn't make sense because it would primarily have been for scraping data. No financial incentive for them to support this.

The dream today is that agents might prefer your website due to it's MCP and actually purchase something. i.e. 'hey claude go find me the cheapest places I can buy these headphones'

Pretty annoying user experience trying to test out the model on the official Qwen site. You get an immediate popup with "Security Notice: For security purposes, please add a payment method" that blocks you from clicking anywhere.

Honestly if you're launching a new model, you should probably just make it free for the 24 hours after launch so people can test it out. Otherwise just wait till it's on openrouter.

I've just been reading The Making of Prince of Persia (highly recommend [1]) and it's amazing how the journal feels exactly like today's startup scene. But everything is dated 1989.

[1] https://press.stripe.com/the-making-of-prince-of-persia

My lord people accept a little bit of whimsey in your lives.

TBH I was a retool power user for years, but I feel like they completely missed the vibe code train.

It's so much more annoying to go drag boxes around and type out little JS snippits in every chart as opposed to just telling claude to make the chart.

FYI we've built this internally and it was a pain, so I was excited to check out what you've built to see if it was a good fit for us.

But your landing page + video was just slop city. Removes 100% of credibility for me.

I know the advice is launch early and often, but honestly I would just point people to the github until you have something better.

Seriously. We got 116 github dependabot alerts this week. Half of them for dev dependencies.

I feel like it's also been overrun by a lot of spam. As someone running a company, I get 2-5 unsolicited "vulnerability reports" per week. Half of them are an LLM finding some bad CSS on our framer splash page. The other half I assume are an extortion attempt so we just mark as spam.

Occasionally I see real security researchers on HN complaining that no one takes the disclosure seriously, or that people reply immediately with a cease and desist. But from the receiving end it's just because the spam is unmanageable.

It's cheap at $4/1k, but I'm hesitant to even benchmark this one again since the previous versions were all "98% accurate based on internal benchmarks of 4 pdfs" and ended up falling short of almost everything else on the market [1].

Even in this one, they just report that OlmOCRBench and OmniDocBench have "known limitations" and that's why they report flagship numbers from their internal benchmark.

https://getomni.ai/blog/benchmarking-open-source-models-for-...

The fact that the top comment [1] sparks a thread debating how CORS does / doesn't work is a great argument in favor of the original title.

[1] https://news.ycombinator.com/item?id=48616086

Agreed! If you're a 20+ company in SF you're required to offer commuter benefits (up to $340 / month).

That's usually things like caltrain / muni. But I would definitely sponsor a $300/mo waymo subscription if it was like 20 rides a month.

That's literally not what he said. He's saying the majority of people supporting border controls are not racist, but the vocal minority are the ones who "boast about arrests" / "make up stories about crime or eating cats and dogs"

I'd wager that even if you didn't nerd out on computer architecture, just living through progression of CDs -> mp3s -> ipods -> streaming gives kids a better grounding than the iPad is where music comes from they have today

As someone who grew up in the 90's, I think seeing the live progression of tech was really helpful for my own understanding. For instance we saw:

- CDs moving to Mp3s moving to the ipod and finally streaming

- Games moving from 8bit to early 3d graphics to where they are today

- Family computer moving to laptops and eventually to ipads

- Landlines to early cell phones to the iphone today

All of these experiences helped ground the core principals behind this technology. And the pace of these transformations (while rapid) was still something you could keep up with. Everything was built on the same principals.

But today kids go from zero to iPad + AI generated tiktoks by time they turn 2. Sure parents can try to hide the tech, but it doesn't change the fact that it's out there and available as soon as they enter school.

Maybe I'm overindexing on my childhood, but I would love to recreate some abridged history of this for my kids. I think seeing the building blocks helps build a much more healthy relationship with technology.

Sure the consumer won't consume 10x more, but they're still going to reach for the better products.

And let's say that work is correlated with quality. Company A wants to spend 10x less time working, while Company B works 10x more. Company B therefore has a better product than Company A, so eventually Company A goes away. The consumer still consumed the same amount, but they switched to the better product.

Yea it's always been competition that's the issue. Greed too. But complacency is really difficult as a business owner.

In the world where someone can take your cake by working 25% more hours, it's always going to happen.

> If a junior makes a mistake and it will not be caught in time they will automatically learn.

I think this sentiment applies well to junior software engineers (with mentorship). But imagine the much larger swaths of entry level employees in operations, support, or sales functions. When you have a 400 person team with 20% annual turnover (since people move in / out of entry level jobs frequently), the management + training + monitoring becomes a huge challenge.

I think the typical HN sentiment of "llms aren't deterministic" fails to take into account how non-deterministic giant groups of people are. Every group of 10 people typically needs a manager. And every 10 managers needs another manager. By comparison the engineering work on dialing in your LLM guardrails feels pretty worthwhile.

Just dropping here because it's an excellent read on US port automation

https://www.construction-physics.com/p/do-us-ports-need-more...

Yea worth it. The original implementation ended up being the most complex, and also not a great UX. But I didn't really get it was a worse UX until I built it and tested it out a bit.

And I wasn't attached to that complex implementation in the way I would be if I architected it from scratch, so it was easy to move on.