what's wrong with the web app

If it was a company it'd have failed already.

> The program, called the Disability Case Processing System, or DCPS, was designed to improve case processing and enhance customer service. But six years and $288 million later the program has “delivered limited functionality and faced schedule delays as well as increasing stakeholder concerns

https://fedscoop.com/problem-project-threatens-progress-soci...

And that's just one instance.

Can you imagine raising $288 million from VCs for a software application while delivering so little?

But taxpayer money? Free and easy money to keep wasting coz no one cares. Tragedy of the commons.

For the main system they're also using COBOL, which has no Date data type, causing issues even in 2025.

One big difference is management control. People feel that government administered services tend to have poor management and citizen services more often than not. One big example is the DMV since almost every has experience dealing with it, long queue times are almost universal because no one gives a crap and it's very hard to fire a government employee. Or the passport issuance, or applying for permits. Or unemployment benefits, the list goes on and on.

Imagine if the DMV and passport services had even the possibility of competition like a private company has. You bet all of a sudden the service would get much faster and better and with fewer mistakes and red tape with the same or fewer number of employees. Or someone would set up a competitor and imagine how many people would even pay extra just to not waste several hours of their time.

It's tax payer money so there is a lot more waste than even at big private companies. For example, the costs to just administer and operate the social security administration(not including any money paid out to recipients) is $15 billion dollars with a big B. There is no incentive for anyone to save the tax payer any money and there would be a huge pushback from govt contractors, unions and employeees. See how much hate DOGE gets for even proposing cuts or higher efficiencies.

Any large IT project in the government in almost any country and at any goverment costs huge amounts while not returning much value if any. Look at the state and costs of local metro stations and trains in almost any city.

Describing dithering as scary is wild

I was initially confused: the article didn't seem to explain how the prompt injection was actually done... was it manipulating hex data of the image into ASCII or some sort of unwanted side effect?

Then I realised it's literally hiding rendered text on the image itself.

Wow.

My guess is that this title could also be written as, “The value of AI projects are being captured by just 5% of companies.”

It’s pretty clear to anyone who’s using this technology that it’s significant. Theres still tons to work out and the exact impact is still unknown. But this cat isn’t going back in the bag.

> Replace junior developers with Claude Code or similar.

I don't know why everyone goes to "replacing". Were a bunch of computer programmers replaced when compilers came out that made writing machine code a lot easier? Of course not, they were more productive and accomplished a lot more, which made them more valuable, not less.

At the risk of being obvious, this seems set up for failure in the same way expecting a human to catch an automated car's mistakes is. Although I assume mistakes here probably don't matter very much.

I am working on a project that uses LLM to pull certain pieces of information from semi-structured documents and then categorize/file them under the correct account. it's about 95% accurate and we haven't even begun to fine tune it. i expect it will require human in the loop checks for the foreseeable future, but even with a human approval of each item, its going to save the clerical staff hundreds of hours per year. There are a lot of opportunities in automating/semi-automating processes like this, basically just information extraction and categorization tasks.