G'day Mattheus,
Was fun to see some of the things I find fascinating causing fascination in others. I'm a bit of a Common Lisp tragic and I also love the APE setup and using llava files to run local LLMs.... Even on tragically old servers!
Hope things go well. Thanks for doing the interview. I'll look into Lone Lisp.
Cheers
I'd really like it if the United States could very soon get rid of or at the very least eventually vote out Donald Trump.
The Nature article points to the MOF Nobel prize winner going to China. I gather this is because science is getting defunded in the US? I think defunding science shows an incredibly short term view and damages the future for the world, starting with America.
I know there are problems with the US democracy that have been in play for decades. There is inequality, there is racism and religious bias... I think throwing out the entire system that largely worked doesn't solve these problems and in fact makes the future worse for our kids. The video above argues that the damage done now to universities will be noticeable in decades time. Even doing a large funding push to universities may not fully repair the damage being done now.
The US is not a perfect place nor a perfect actor, but when governed reasonably I think it is better than China in terms of its aims, philosophies and interactions. Hopefully his work done in China will be shared globally.
I've heard it said that growing crystals in gravity is tricky... If we could do this in space, gravity free, we could get better crystals.
Is that correct? Pardon for asking a bit of a question to your comment. I don't know too much about growing crystals for things like jet engine blades.
Thanks so much for being bold enough to be fairly open about the costs, how you arrange billing and the advantages that's given you.
I've been fooling around with DeepSeek 4 agentically. It's probably not as good as Anthropic offerings, but even those seem to be roiled in politics and strife and DeepSeek 4 is very good IMHO. I'll later try out GLM.
I'm in Australia. The government has set up a "return and earn" scheme to keep aluminium cans, plastic bottles and paper drink cartons out of the waste stream. A laudable project. The money you make from return drink containers is pretty low, $AU 0.1 per container. I've participated to get the rubbish out of natural water streams and to make a nano amount of money on the side.
When I looked at the costs of an app I was getting DeepSeek to help me with, I realised that the several hours I'd spent learning and building had cost something like 8 recycled containers. In my head after doing some DeepSeek stuff, I calculate a "cans per app" metric for myself for fun. I may even setup a simple graph to view my costs that way.
I kind of hope the Anthropics of the world get enough price competition from sources like DeepSeek and GLM to drop their prices significantly. Time will tell.
I'm using the Chinese DeepSeek provider, so everything done there could potentially be taken and used by the CCP... But this is hobbyist learning.
There is probably a market for Deepseek/GLM served from non CCP available servers. I might even look into how hard that would be to setup here.
I also hope that inference focused hardware will come to the fore, reducing energy use and cost. Realistically this will take time though, on the order of years.
Here in Oz, we have community batteries that community members can charge and later draw from. Their electricity prices are competitive. I wonder if someone could setup something like a community battery to run data centres... That way reasonable environmental consideration could be given to inference power generation... This might not work in a market like the US or Europe, but small market size might be an advantage... Who knows.
This comment is not entirely on point with your comment, it circles around and above it looking for lift though.
If you're not doing work that requires your code to stay in home nation data centres, Claude for Deepseek, Deepclaude (https://github.com/aattaran/deepclaude) is a great way to get better at using Claude like tools for software development. It even does a pretty good job of putting together cover letters for job applications...
Using Deepclaude is very much cheaper than using claude... For hobby projects, I've found it useful. A recipe (for cooking) management app I've made took a couple of hours to put together and cost $US 0.5. Claude is far more expensive.
The downsides of Deepclaude for many are:-
- DeepSeek is a Chinese corporation so the Chinese Communist Party may ask for data if it wants it.
- DeepClaude isn't as fast as normal Claude, though it's still pretty fast and I think fast enough (YMMV).
- DeepClaude might not be as optimised for various code issues that Claude may be able to solve more quickly or effectively.
- The same safeguards are probably on DeepSeek, but you won't be "wasting" as much money as you might on using Claude.
Inference focused hardware (https://www.youtube.com/watch?v=nvPqHoVSenE, AI generated speech) may in the medium future cause a large enough cost/energy reduction for LLM tools like Claude to make local LLMs more attractive.
Inference focused hardware would make running Open Source models like DeepSeek on local machines far cheaper and control over safeguards would return to the end user.
Hopefully this leads to a localised LLM provision market where local businesses provide varieties of these "local" LLM services. Here, local could mean on premise through to state or nationally based LLM services. Eventually, government orgs outside of the US may demand this kind of LLM use, in the same way governments legally require data to be stored within national borders for many critical government functions.
A bloke can dream I guess...
...Could affordable inference focused hardware also cause the bottom to fall out of these stock market bending valuations for AI corps and their datacentre obsessions?... Not to mention the societal costs caused by the AI super corps building these data centres. At the moment, they're nearly making a profit... They seem almost like speculative companies... Is that a term?
Having Blaise work on Windows, could be interesting... Though of course only in the long tail of already built Delphi apps... that still have source.... that are reasonably written... that don't rely on now unsupported libraries...
Hmmm... Ah yes. These may be some of the reasons people might not do Delphi as much anymore. Still, Delphi was great for me and helped me get a mortgage and things like that.
The Blaise author is progressing his development along nicely using QBE for the compiler. To me, Blaise looks like a team progressing things nicely, but I think it's mostly the original author.
Blaise could be a neat Pascal compiler with a great deal of old cruft cleaned out and is quite quick.
I get what you're saying about Gemini for coding and it's useful that you mention it.
I wonder though if Google isn't so worried about the viability of their coding AIs and have a longer term view than simply providing coding aids. This might also be indicated by their recent $40B investment in Anthropic, https://www.cnbc.com/2026/04/24/google-to-invest-up-to-40-bi...
I thought that might be the case. I naively wondered.
I'll see if I can understand the paper :-)
Hope the paper gets lots of references and the technique gets a lot of use to save power and time.
There's been several potential big changes for LLM inference efficiency over the last few months. There's been Attention Sequencing (I think it's called..?) Turbo Quant and this one.
There are multitudes of neuro research projects using animals as test subjects that could save a lot of money using a tool like this. If the research project is able to use a device like this rather than a human verified device. From my limited knowledge, these projects do their best to be very kind to the animals being tested. A device like this wouldn't need to be enclosed within the animal, so less risk of harm right away.
In Africa, clever locals built a humidy crib from car parts. It ended up not working as a product, but a great idea. In the link below, its celebrated as a commercial failure to learn from.
I think EEGs (Electro EncepheloGraph) can produce a far more detailed, brain related view of what an ECG (Electro CardioGraphs) can produce. An EEG can of course look into many other brain related functions and issues.
Creating a low cost version of an EEG will hopefully at least provide some thoughts to the engineers of commercial EEGs.
Commercial, medically verified devices are tied down in many ways...
- Full checks of the software and hardware design,
- backwards interoperability and compatibility for devices and their connections over their lifespan
- A full medically based software/hardware quality check,
- Providing very detailed documentation,
- doing a full test cycle around every device,
- Interacting with doctors and health experts to fully characterise the domain and typical device use. This is great to do as an engineer, but is expensive for the company :-)
- Older (often slower) more fully field verified/trusted chips for any logic are used, since they have a large measure of reliability, reducing the risk you'd get from new products.
The list above is from memory. Engineering around devices like this become part of the culture of the company. Each region (US, Europe, Oceania) have their own requirements and levels of completeness. The big market for any medical product is the US. It's FDA in the past has been the most important regulatory body to satisfy to allow access to the US market. Several other markets use the FDA as a base for their own standards too.
This process takes several years and millions to complete. Its a very necessary step. Think of the Therac 25, https://en.wikipedia.org/wiki/Therac-25 . The mistakes in this device design caused fatalities.
I used to work for Cochlear. Their devices have had some issues, but overall I think they have a very good record. The verification/quality checks outlined above are super necessary.
Anther Oz company Telectronics, built many pacemakers and unfortunately made a design mistake in the choice of material for a lead feeding the part of the device that fed the therapeutic current to the patient. After time, the lead cracked. Failure of physical integrity in a fully enclosed medical device is quite a bad failure mode. I believe The company lost 75% of its share price overnight when the issue was reported. https://en.wikipedia.org/wiki/Telectronics
When this device becomes fully wireless, its risk will drop somewhat.
This project is awesome also as a teaching tool for people wanting to join companies that make medical products. They'll be able to look over a modern design and discover what's involved.
Please jump in and make any additions or corrections to the above.