sunpazed

I’m a vintage calculator tragic! I love what the C43/47 communities are doing in reviving RPL and RPN for a modern audience. What’s even more amazing, is that the DB48x calculator is a singular dev effort by the talented Christophe de Dinechin (I’ve contributed briefly to his code).

If you’re curious to try out RPN/RPL without buying hardware, then both the R47 and DB48x calculators are available as apps on the iOS App Store.

Those of you who want to go deeper down the rabbit hole, I strongly recommend you check out Calculator Culture — he reviews both modern and vintage devices; https://www.youtube.com/@CalculatorCulture

This is really impressive. At these speeds, it’s possible to run agents with multi-tool turns within seconds. Consider it a feature rich, “non-deterministic API” for your platform or business.

Thanks so much for delivering on this model. It’s great as a draft model for speculative decoding. Keep up the great work!!

Don’t have enough ram for this model, however the smaller 20B model runs nice and fast on my MacBook and is reasonably good for my use-cases. Pity that function calling is still broken with llama.cpp

In my testing, tokens per sec is half the speed of the GPU, however the power usage is 10x less — 2 watts ANE vs 20 watts GPU on my M4 Pro.

The key benefit is significant lower power usage. Benchmarked llama3.2-1B on my machines; M1 Max (47t/s, ~1.8 watts), M4 Pro (62t/s, ~2.8 watts). The GPU is twice as fast (even faster on the Max), but draws much more power (~20 watts) vs the ANE.

Also the ANE models are limited to 512 tokens of context, so unlikely yet to use these in production.

Love this! The C64 introduced me to the world of computers as a kid. I still have that almost 40 year old machine in my collection, but I’m weary of failure every time I turn it on. This is somewhat better than the MiSTer as I can use physical peripherals with it. Great work!

Let’s remind ourselves that MCP was announced to the world in November 2024, only 4 short months ago. The RFC is actively being worked on and evolving.

While I’m a fan, we’re not using MCP for any production workloads for these very reasons.

Authentication, session management, etc, should be handled outside of the standard, and outside of the LLM flow entirely.

I recently mused on these here; https://github.com/sunpazed/agent-mcp/blob/master/mcp-what-i...

Funny to see you here. I’m from operata.io (also a Melbourne based startup) and would see your website (operatr.io) when I would mis-spell mine!

I had the same frustration and wanted to see "under the hood", so I coded up this little agent tool to play with MCP (sse and stdio), https://github.com/sunpazed/agent-mcp

I really is just json-rpc 2.0 under the hood, either piped to stdio or POSTed over http.

Great stuff! I used both LiteLLM and Langfuse for a recent project at work — both tools were instrumental in development and deployment. Looking forward to seeing how this collaboration develops.

Even though I have a reasonable intuition and understanding of how a LLM works, I am still awe-struck each and every time I use one. The fact that I have a junior developer at my convenience is a huge efficiency gain. I’ve been able to automate the rudimentary elements and focus my time on the stuff that counts.

Operata | Data Engineer, Business Dev + others | REMOTE (AU & US) | Full-time

Operata is building the world's first observability platform for Cloud Contact Centers. We're a VC-backed, small 13 person team — and have grown significantly in the last year. Operata is an AWS Select Technology Partner, and our customers include leading insurers, telcos, banks and managed service providers. We have serious traction and are scaling up!

https://operata.com/careers

We're hiring a few roles:

Data Engineer — Looking for expertise in ETL/ELT processes, data lake-house platforms, data modelling and data pipeline development. Proficiency with real-time data processing platforms such as Apache Flink, Spark, Airflow, or DataBricks is a plus.

Data Analyst — Looking to improve our data culture, establish an analytics practice, and level-up the capability of our Observability product.

Director, Solutions Engineering + Director, Business Development — US (Remote), PT or CT Time Zones. Looking for experience in technical sales, solutions architecture, sales engineering or enterprise sales in cloud or communications related software products.

Check out the webpage, or DM me for more.

This is an unfair and snide take on Woz. We are literally standing on the shoulders of giants — Woz happens to be one of them.

I have a few vintage programmable HP calculators that implement CORDIC on their 4-bit CPUs. You can also program them to calculate the Taylor expansion of sin() as another approximate method.

If you liked this, it’s worth reading Donald Knuth seminal “The Art of Computer Programming” which explains a number of mathematical algorithms by example.

I had a .beats watch and found it never really achieved its aim of abstracting timezones — I’d wake up at @895 and head to bed at @580. I’d cross the 1000 boundary each day, which did nothing but highlight the international timezone difference.

Interestingly outside of games and consoles there are “emulators” that run classic HP calculators in the store. They emulate the HP Nut processor which runs small ROM dumps.

https://longform.asmartbear.com/logo/

The “European hacker-culture demo” from Ezra’s Amiga is likely State Of The Art or 9 Fingers from Spaceballs — both really cool scene demos:

https://youtu.be/89wq5EoXy-0

https://youtu.be/n4M7e79XTYk

These easily run on an Amiga 500 ES from the 80s, and are stored on a single floppy disk.

After our fancy fridge died, we purchased a “dumb” Mitsubishi Electric WX Series fridge. No LCD panels, no wifi, no time.

It’s tiny at only 470L, however it’s designed for maximum storage in small Japanese apartments, so we can fit heaps of groceries and frozens. It also has a chilled compartment specifically for sushi meat!

Also, 10 year warranty on all parts.

A coffee table book about Keyboards that is also a functioning Keyboard /s

This technology has been around for a while on a number of different platforms. Wizzard VHS on the C64 predates this from about 1987-88

https://www.lemon64.com/forum/viewtopic.php?t=61209

“The tragedy of the Luddites is not the fact that they failed to stop industrialization so much as the way in which they failed. Human rebellion proved inadequate against the pull of technological advancement.”

https://www.newyorker.com/books/page-turner/rethinking-the-l...

The Game Boy is by far my most favourite console of all time. I’m currently writing a game for it in ASM (a pinball-dungeon crawler). The hardware is simple to understand, and the restrictions require you to be creative.

I understand where this is coming from. I had a short 4 year stint in a mid-size tech company, and led a small 16 person team. We managed to build and scale a fairly successful product, to the surprise of the much larger and well resourced teams.

Unfortunately, success attracts attention from management, and we found ourselves mired in further layers of bureaucracy. Those who had not really owned products or shipped anything significant began to call the shots.

I’d agree — but would argue affordable for a sponsored dissertation program with 3 research students and an associate professor. They’re actually still training it!

While LLM projects typically require an exorbitant amount of resources, it is important to remind ourselves that research does not need to assemble full-fledged massively expensive systems in order to have impact.

Check out TinyLlama; https://github.com/jzhang38/TinyLlama

Four research students from the Singapore University of Technology and Design are pretraining a 1.1B Llama model on 3 trillion tokens using a handful of A100's.

They're also providing the source code, training data, and fine-tuned checkpoints for anyone to run.

Rolex have innovated consistently with their movements and oyster cases, and their manufacturing quality is amazing. Their business practices are questionable though, as the article highlights. I can walk into an AD today and walk out with a Grand Seiko Snowflake. Contrast this with a Rolex AD who will put me on a “waiting list”, even though I’ve purchased from them before and own a few Rolex watches.

I would argue that they used to be. Back in the 70s a Rolex was seen and marketed as more of a “tool watch” than a luxury item.