jochem9

I listen to radio a lot. NTS, FIP, Concertzender. It's a great way to discover new music and have some human presence.

Recently a DJ with excellent music taste and a deep radio-perfect voice died. I was devastated when I got the news. The two remaining radio shows in his slot were something I'll remember forever. One unbelievably raw, as it was only two days since he died, the other beautiful, as everyone came with music and stories to remember him.

You don't get this connection over streaming.

RIP Bob Rusche

Right now the article reads as "AI can play doctor if you give MRI scans".

If the author would actually go for a second opinion (maybe bring along the AI to let it explain it's findings), then the article could read as "AI did MRI analysis and proved my doctor wrong" (or: "AI did MRI analysis and failed").

There was an article on HN a few weeks ago where someone detailed how they managed to get an old datacenter GPU to run in their consumer PC, getting decent performance with qwen. He spent something like $200 on the GPU (second hand of course).

So yeah, I think models on local hardware will be quite common soon among the tech savvy (such as people creating software).

Wow, totally forgot I've used this rss reader in the past. Happy to still it's still rocking and stronger than ever!

Ps: love the statement "it's like podcast, but for reading".

First iPhone didn't have support for 3rd party apps. Steve Jobs even explicitly spoke about wanting to have all 3rd party things run in the browser.

Only when jailbreaking and custom apps got very successful, Apple introduced official app support and the appstore.

So maybe now the advice could be to build a passable mvp and then use that to figure out the product-market fit.

It's a lot easier to think about it and discuss it with others when the product is tangible instead of just an idea. Biggest challenges then are being able to still diverge (broaden/pivot the product idea) and killing functionality that doesn't work out afterall.

This is the third HN post I read on this topic. Everytime the same tweet (or whatever it's called for mastodon/bluesky/etc). Did anyone actually debug the issue?

Was it caused by poorly generated code, or was it caused a genuine (security) fix that accidentally caused it (potentially even in a way a human would to)?

AI has no taste, so I suspect the labs just gave it a bunch of decent looking boilerplate as preferred style.

When you bring your own ideas you can get AI to dev pretty nice looking non-generic stuff.

I work with stakeholders that come from different backgrounds (different countries, non-engineers). No way that we can get aligned using just text. Or if we try, it will take a tremendous amount of back and forth, annoying everyone in the process.

I'd much rather talk for 30~60 mins and get everything hashed out. It also allows you to build rapport, so next time it will be much easier to do something together again.

Yes. As someone else points out: the techniques for this exist (and have existed for decades). It was never worth it to fully pursuit it, especially for more messy human-heavy processes.

Now with AI you can get way more detailed (AI can interview humans, and you can do it in a format that doesn't feel like an interview - e.g by 'simply' having AI be a fly on the wall). AI can make sense of messy inputs and then you can present that to people who know the process, who can easily point out flaws. The flows/maps will not be perfect, but you can always get more detailed once you bump into the limitations of imperfection.

And absolutely do business leaders want maps like this. They often have no idea what exactly is happening in their org. Or they have a gut feel that there are massive inefficiencies, but they cannot get the insights into where they are because there are too many moving parts that do not effectively talk to each other (silos, politics, etc).

I suggest to hope less.

The world order we knew is upended and it's likely to stay that way. Better to spend energy shaping that into something that is acceptable, then hoping it will all go away once DT is gone.

You can put a limit on token spend and provide training (and even pre-configured workflows) on how to limit token spend.

Like the other commenter said: cloud spend can also spin out of control if you don't pay attention, yet we've found ways to keep it under control (training, guardrails, limits, transparancy).

Ask them to share their prompt instead.

Calls them out on their AI bs and gives a way forward to share what they actually thought.

I am working with agentic AI on industrial manufacturing data. The speed at which you can get insights and dig into all kinds of rabbit holes is insane. It's just as easy to compare plants so you can make strategic decisions on budget allocation as it is to do root cause analysis on why during a given shift there were so many breakdowns.

And this happens with a natural language interface, instead of Excel (although people of course still want an export to Excel button) or worse: by having to go to the BI analyst, have them change a dashboard and after waiting for a few weeks hope they give you what you want...

Yes, you need to structure your data well. Especially metadata/defintions and accessibility - which is not cleaning up ETL pipelines, although that will help. And obviously have lots of relevant data available already (which was my job before this).

From my experience: fully automated marketing budget allocation... Doubt it. Time to insight reduced >10x? For sure.

The exact outcome depends on the country's specific laws, but generally speaking, there is protection.

In the Netherlands you get up two years paid sick leave, before your employer can fire you. If you are sick enough to not be able to work again, you'll get 75% of your last earned wage from the government.

The tricky situation is when you are considered partially unable to work. You'll get time to find a suitable job, but your benefits drop over time. Finding a job in that situation is sometimes very difficult. It's possible to fall between the cracks.

In any case, the intent is to make sure people do not get unfairly hurt by life events outside their control.

I used to do lots of data engineering in Python, then started doing all kinds of engineering in primarily Go.

The Go ecosystem for data is very limited. There are no widely supported dataframe libraries (like the og pandas and the newer polars written in rust and also available as a crate). Very few data science libs, a few decent gen AI libraries, but not as popular as their Python cousins.

Most of the work I do now is streaming data and very small batches. For that Go is amazing. I don't need dataframes to transform a json, combine it with a bunch of other data and write to a database. I just need to write that logic and make it go fast. Very easy in Go.

If you're at a stage where code doesn't compile and doesn't do what you want, then that's really easy to improve.

- let the llm (agent) compile the code and iterate until it does.

- state your intent more clearly.

This implies that the government respects the rules or at the very least pretends to do so. To me it's pretty clear that the US federal government has moved beyond that.

One does not rule out the other. In the end it's nerds messing with hardware.

Lots of computer culture is rooted in anarchism, anti-capitalism and a fight for fairness. E.g. early internet culture, the open source community.

Imo it's very nice to see explicit anti-capitalist movements within tech, because the other side of tech is so completely over the top capitalist.

To me these all sound pretty plausible?

You have browser extensions that allow AI to control it. There are apps that create a virtual microphone for text to speech. All pretty easy to get going if you allow an AI to control a computer.

Generating designs is also easy. Just use nano banana or something similar.

I'm more curious about the quality of the output. I've seen some ugly stuff. The article also includes one example (the outdoor seating design).

I'm also curious about the reaction of people to AI running the shop. Do some people decide not to deal with it? Do some people abuse it? In the article there seem to be only positive examples of human-AI interactions.

Says a lot about the state of society when parenting is outsourced to technology, so that the parents can be further enslaved (because almost no one chooses to work two jobs).

EU severely reducing its fossil fuel imports from Russia in 2022 cut down natural gas usage by 17% and overall energy consumption by 3%. So yeah, increased price due to scarcity help a lot in shifting around the energy mix.

It's a bit shit that hits poorer people relatively more than richer people. Governments can reduce this impact by subsidizing sustainable alternatives (like heat pumps). It's still leading to inequality (unless you give more subsidy to the poor), but at least overall people will hopefully benefit.

I'm on a very similar train. You cannot dump all the data into an LLM (for many reasons) and we also already have clearly defined rules that an LLM doesn't have to figure out.

So keep organizing data (LLM powered, of course), so that you can query data as usual (multi modal, so not just graphs, but also time series, relational, etc). Feed that to deterministic computations. Let an LLM reason about the outcomes.

Give the LLM the freedom to orchestrate the retrieval and computations. Make sure the way it orchestrates it is auditable.

The key thing I want to achieve is beyond this system: I want to uncover hidden things in the system (missing in the ontology, computations, etc) and propose to add these. This will effectively give you a generic approach to create ever evolving systems aliging with reality while being fully auditable.

I'm also doing openspec for a few months now and it's really good if you invest enough in the specs (in the beginning I skimmed over much, now I pay attention to all details and fix anything that's wrong or where I see a gap).

The one thing I like that OP brings is to tie specs and code together. The openspec flow does help a lot in keeping code synced with specs, but when a spec changes, AI needs to find the relevant code to change it. It's pretty easy to miss something in large codebase (especially when there is lots of legacy stuff).

Being able to search for numbered spec tags to find relevant bits of code makes it much more likely to find what needs to be changed (and probably with less token use too).

Not the original commenter, but I share the same sentiment.

The harm done is that there is less human interaction outside one's bubble. Before deliver-anything-at-home you were forced to be in environments where you'd interact with people you'd normally not see.

At some level you're also building a connection if you're interacting with the same service employee or bump into a neighbor.

Strong communities is a massive boost to public health. It reduces loneliness and prevents all kinds of deceases (heart stuff, dementia, etc). It also builds a safety net so people don't have to rely so much on official healthcare.

Now obviously just some food delivery is not destroying community by itself. There's probably worse offenders out there. However, society becomes more and more parallel and robots do contribute to it. Unfortunately it's not something that's discussed a lot.

Most products you consume are probably not software. Pretty much all products you consume are created by companies that use software.

If those companies don't keep up with software, they might not have a competitive edge that their competitors who are keeping up gain.

Yes.

Your message needs to find other people. The how is irrelevant, it just shapes and transmits it.

Oh, I thought it was opium.

This is the crazy part with LLMs. It knows much more than you as a single user will ever realize, as it only shows the part that matches with what you put in.

I was building a tool to do exploratory data analysis. The data is manufacturing stuff (data from 10s of factories, having low level sensor data, human enrichments, all the way up to pre-agregated OEE performance KPIs). I didn't even have to give it any documentation on how the factories work - it just knew from the data what it was dealing with and it is very accurate to the extent I can evaluate. People who actually know the domain are raving about it.

You could build (code, if you really want) tools to ease the review. Of course we already have many tools to do this, but with LLMs you can use their stochastic behavior to discover unexpected problems (something a deterministic solution never can). The author also talks about this when talking about the security review (something I rarely did in the past, but also do now and it has really improved the security posture of my systems).

You can also setup way more elaborate verification systems. Don't just do a static analyis of the code, but actually deploy it and let the LLM hammer at it with all kinds of creative paths. Then let it debug why it's broken. It's relentless at debugging - I've found issues in external tools I normally would've let go (maybe created an issue for), that I can now debug and even propose a fix for, without much effort from my side.

So yeah, I agree that the boring part has become the more important part right now (speccing well and letting it build what you want is pretty much solved), but let's then automate that. Because if anything, that's what I love about this job: I get to automate work, so that my users (often myself) can be lazy and focus on stuff that's more valuable/enjoyable/satisfying.