chromaton

FWIW, Reuters is saying the total losses are in the range of 10's of millions of euros across all stores.

I addition to the technical knowledge needed to make this work well, there are misaligned incentives between the store managers and the company as a whole.

One particular example: store managers want a very cold, very narrow temperature deadband on their frozen items. This gives the customers a nice experience. They're less likely to complain about melty ice cream. But this type of setup leads to more energy expense, and failures on the compressors due to excessive cycling.

Grocery is a very low margin business, and they look for savings wherever they can.

They weren't able to set up to work as efficiently.

Some people have to learn the hard way, I guess.

> why it needs to be fiddled with at any significant frequency

Weather changes, equipment performs differently as it wears, changes in customer behavior, different items being stocked, electricity cost updates...

As I mentioned in the thread regarding the previous attacks, I worked in a closely related field back in 2008-2009.

One thing I see repeatedly brought up here in this discussion is "Why can't we just set the refrigerator temperature with a knob and walk away?"

The problem is that these refrigerator cases are not as simple as your refrigerator at home. As a simple example, many systems allow you have control over the deadband: the exact low temperature at which the compressor turns off and the high temperature at which the compressor turns back on. You also have control over defrost cycles, compressor pressure and many other things (I worked with this a long while ago, and I don't remember all the details).

Take a look at some of these manuals, for example: https://www.hillphoenix.com/resource/installation-manuals-2/... As far as I can tell, those manuals just cover the refrigerator cases, and not the outside compressors and heat exchangers, which are another layer to deal with.

This is not something that a non-expert supermarket employee would be able to handle well. The purpose of all these settings is optimize around energy usage, maintenance costs, customer experience, etc. Even if you only save $500/store/year, you can see that it really adds up over a chain of 1000 stores.

That's why this equipment is remotely accessible: it's not like your fridge at home and requires a lot more knowledge to get the most out of it. Of course, that does not excuse poor security implementations.

It worked for me, and I'm getting about 6% faster training on nanochat with my 3060 card. Thanks for this!

However, it required a few extra steps, as the repo is set up for Linux kernel 6.x and I'm on 5.15. Claude Code was able to get me past the hurdles, including code modifications and startup scripts.

Right now, I'm clearly power limited at 170 watts and I have no desire to brick my card or set it on fire.

Grocery is very low-margin, and if a new system can save $500 on electricity and maintenance per year per store, and you've got 1000 stores, the technology justifies its cost very quickly.

Well, many of them were still on dial-up modems in 2008.

Saving on labor is the purpose of most technologies since the pointy stick.

I worked in the field of control systems for supermarkets back in 2008.

These systems are internet connected and remotely controllable because the managers need to be able to see shifts in usage over time and take action to save electricity and provide better food storage. One example might be that due to shifting customer patterns and products, you might want a certain refrigerator case to be kept colder with a narrower temperature range. Another example is for Demand Response: when a brownout may occur, the utility companies want people to lower their load, which, in the case of supermarkets, means turning off lights, increasing the thermostat on the AC, turning off dehumidifiers, etc.

One notable story from that time: this was during the 2008 economic crash, and the company went from hiring to laying off within a week. A group of technicians were working on a refrigerator compressor system and were told by security to stop what they were doing and immediately exit the building. They tried to explain that they were running a special test and needed to shut things down before leaving, but the security guys weren't having it and made them leave right away. The test ended up running overnight, and the managers came in the next morning to over $100,000 worth of damage to the refrigeration equipment.

A classic satirical SF short story from Stanislaw Lem, containing a satirical look at AI, sexbots, and transhumanism.

I did something very similar last year, but with programming languages that were REALLY out of distribution; they were generated specifically for the benchmark. I call it TiānshūBench (天书Bench): https://jeepytea.github.io/general/introduction/2025/05/29/t...

Some models were OK at solving very simple problems, but nearly all of them would, for example, hallucinate control structures that did not exist in the target language.

Historically, the cycle has been requirements -> code -> test, but with coding becoming much faster, the bottlenecks have changed. That's one of the reasons I've been working on Spark Runner to help automate testing for web apps: https://https://github.com/simonarthur/spark-runner

I've recently found that my ability to add new features and squash bugs has outpaced my ability to do full end-to-end tests. To help with this, I created Spark Runner for automated website testing. It will create and execute a plan for tasks you give it in plain text like "add an item to the shopping cart" or you just point it at your front end code and have Spark Runner create the tests for you. It also makes nice reports telling you what's working and what's not.

New project, so feedback is welcome.

TFA seems to be big on mathematical proof of correctness, but how do you ever know you're proving the right thing?

Lisp has been around for 65 years (not 50 as in the author believes), and is one of the very first high-level programming languages. If it was as great as its advocates say, surely it would have taken over the world by now. But it hasn't, and advocates like PG and this article author don't understand why or take any lessons from that.

Moravec strikes again.

For my benchmarking suite, it turns out that it's about 1/5 the price of Claude Sonnet 4.1, with roughly comparable results.

If you're looking for free API access, Google offers access to Gemini for free, including for gemini-2.5-pro with thinking turned on. The limit is... quite high, as I'm running some benchmarking and haven't hit the limit yet.

Open weight models like DeepSeek R1 and GPT-OSS are also made available with free API access from various inference providers and hardware manufacturers.

This has been available (20b version, I'm guessing) for the past couple of days as "Horizon Alpha" on Openrouter. My benchmarking runs with TianshuBench for coding and fluid intelligence were rate limited, but the initial results show worse results that DeepSeek R1 and Kimi K2.

Current AI systems don't have a great ability to take instructions or information about the state of the world and produce new output based upon that. Benchmarks that emphasize this ability help greatly in progress toward AGI.

Yes, it would be fantastic to have more languages to test off of. I picked the base language I did (Mamba) because it was easy to modify and integrate into Python.

Generating the problems: I just thought up a few simple things that the computer might be able to do. In the future, I hope to expand to more complex problems, based upon common business situations: reading CSVs, parsing data, etc. I'll probably add new tests once I get multi-shot and reliability working correctly.

New base programming languages would be great, but what would be even better is some sort of meta-language where many features can be turned on or off, rather than just scrambling the keywords like I do now.

I did some vibe testing with a current frontier model, and it gets quite confused and keeps insisting that there's a control structure that definitely doesn't exist in the TiānshūBench language with seed=1.

I find that these books have to be read by the right person at the right time. Think and Grow Rich by Napoleon Hill did nothing for me when I first was exposed to it, but later on, helped me greatly.

BTW, the business book that helped me the most is barely known: Making Money is Killing Your Business by Chuck Blakeman.