They show this off, but artificial analysis contradicts the statement. Fable 5 cost $3.14 per task, while 5.1 cost $3.69 -- around a 15% jump in pricing.

https://artificialanalysis.ai/

These, IMO, are marginal improvements for a more expensive model. I stopped using Claude ~3 months back; its outputs are too jargoned, it makes architectural decisions that are not right, and it's incredibly pricey for what it is. Each decision it makes, it acts as if a problem as major as world hunger has been solved. And the overly verbose code comments, strange commit descriptions, duplicate code, and slop it generates -- which I know is not specific to Fable -- is just too much for me.

I found the best is to use something like Deepseek V4 Flash -- with a fast TPS provider -- and work on the code myself. For agentic work with computer use, GLM 5.3 flash with Hermes Desktop works well.

https://artificialanalysis.ai/

The costs in this article are understated for products like ours. This is not just for packaging -- they have separate compliance schemes for electronics (WEEE). If selling products with batteries, you must comply with separate battery regulations. If you pass a certain threshold or contain specific chemicals within your product, you must also comply with the REACH regulation. They work almost exactly the same the article describes, but often with higher costs.

Additionally, you must have CE certification for your hardware product. There is no threshold. This means going to a lab and testing RF capabilities, and for wireless / Bluetooth devices it can cost $1,500-$8,000. You can self-certify but it is risky.

We pay 800-2,000 EUR per country annually for the authorized representative (AR) and producer responsibility organization (PRO) fees. The AR's task is to hold some documents and provide them to the government when requested -- but that's it. We can do this for free. There's no reason the costs should be this high -- and I see minimal reason why they should exist in the first place. But we must comply.

It is important to understand which regulations and which fees to pay for the product. Therefore it is important to use something like PRONEXA -- a "One-Stop-Shop for simple compliance management".

With them, we paid ~800-1,000 EUR per country as a one-off registration fee. Then we pay 975 EUR annually per country to keep compliance. Note that these costs are ONLY to PRONEXA, not to each AR or PRO. They are effectively the middleman between us and the ARs / PROs.

The fee stability is quite poor, too. We paid the 2025 annual fees roughly a month before year-end and were charged another full annual fee in February 2026. In November 2025, we paid ~500 SEK for Sweden, and in Feb 2026, we were invoiced ~15,000 SEK for that same line.

With all of this burden we only pay the countries ~0.5-50 EUR annually for the actual environmental fees. This only makes sense for the government if that number jumps up to the thousands.

I think this contributes to lack of innovation in Europe. Small startups can't ship there, and you are limited to large companies overseas. The fix is dead simple -- just introduce a threshold, or charge a recycling fee upfront with each parcel coming in, depending on the HS code. What they did with IOSS is great. You register with one country, and all parcels under 150 EUR can be sent to all of Europe while being VAT compliant. They could implement the same thing for WEEE, battery regulations, packaging, and REACH. Register with one country and cover the rest.

We are barely break-even on some countries, and others we are totally underwater. We decided to do it as a future scaling opportunity, but this first year is quite tough. We are debating if we should continue selling to Europe at this stage.

Did Anthropic or OpenAI fail to sandbox their models again?

https://downdetector.com/

Comments URL: https://news.ycombinator.com/item?id=49307533

Points: 2

# Comments: 0

The conundrum which tricks me though - is this a net negative or a positive? If humans are less intelligent, but their output is 2-3 times more intelligent (with AI), what's the result? At what point do we, as humans, stop comprehending anything and give all intelligent work to the neural nets?

And if that does happen, could we live in a society where no work, or at least a significantly less amount of work, is needed? To me, it seems like a dystopian net positive.

It might seem far-fetched to ask these, but I think these questions are getting more prevalent by the day.

That is, of course, unless they develop their own hardware specifically to run this open model. But, that does ruin the point of open models.

Because of this, I think it might not be possible to have AI *only* open-weight; major players like OpenAI, Anthropic, Google will likely stay for good, with better models than open-source versions.

I think it might look something like Photoshop & GIMP, with Photoshop being a frontier lab, and GIMP being the open-weight model. GIMP is decent for many different image editing workflows, but Photoshop is just better.

I would definitely prefer to have an open-weight model better than frontier labs'. Though I don't think it's possible.

> Restricts model to large corporations

> Release information about how Fable / Mythos 5 is stronger than Mythos Preview, give access to every user for a limited time via subscriptions

> Users jailbreak model

> U.S. suspends Fable / Mythos use

Who didn't see this coming?

I wonder what this means for the future of AI models. Either we'll see worse guardrails than what was there for Fable 5 (for me, it was a unusable at times), or the models just stop getting better from here.

I think it's that the guardrails will be more strict, which is unfortunately not good news.

While I don't want to sound overly pessimistic, the models are improving at a rapid rate. If asked ~3 years ago where the state of the models are today, it would sound like sci-fi if answered, "the models are creating full MVP apps in ~30 minutes with one prompt".

The hurdles the models are facing now, like reducing hallucination rates, ensuring compliance, and keeping a clean codebase, do not seem far away from being resolved IMO. Fetching specific information is already partially done with various MCP servers / RAG.

I am, of course, a bit worried about the future of software engineers. If these quirks are resolved, where do their professions fit in the industry? Delegating tasks to the AI model? Unfortunately, this does not require years of expertise, which is a double-edged sword. Reviewing AI's output? Ask it to explain each line not understood.

I think we will see more waves of larger layoffs, similar to how human computers were replaced by digital computers. To some, doing complex mathematical calculations mentally is a fun task / challenge, but it is ultimately significantly slower and more error-prone than calculating with a computer. In the same way, I think hand-crafting code will be seen as a fun "challenge" and AI will be seen as the "modern-day calculator".

I definitely agree with your statement -- it's unlikely the price to performance and quality ratio on Macbooks are going to be outperformed soon.