smcleod

5 ms·

61.
by smcleod 4mo ago
Yikes twice the price of a PS5 in Australia! I will still be buying one though. I'm looking forward to moving away from Sony after having a PlayStation in my living room for 16+ years. I really like Steam's ethics / how they
62.
by smcleod 4mo ago
Similar conclusion I've come to. For all the complex RAG and many SaaS startups selling knowledge engines - what we keep falling back to is simple a wiki (markdown) + simple tools to solve problems and help answer questions programmati
63.
by smcleod 4mo ago
Thought that might be the case, thanks for taking the time to explain!
64.
by smcleod 4mo ago
Oh right, that's not so bad. Isn't that just being part of modern society? It would be nice to opt to never be recorded but also, it's outside.
65.
by smcleod 4mo ago
Pretty sure that's a violation of fundamental human rights as it's your place of living. Surely that can't be legal, even in the US can it?
66.
by smcleod 4mo ago
Looks nice. Is the source available somewhere? If not, I'm genuinely interested to know what the incentive to release it as free but closed source is? Does that mean it might go paid at some point?
67.
by smcleod 4mo ago
I've never seen any developers really use a windows terminal before, if you're on windows wouldn't you be using the WSL or whatever if you're doing development?
68.
by smcleod 4mo ago
Shows how different countries are. In Australia the vast majority of people seem to have an iPhone, with Android being the outlier (at least in the major cities, I can't speak to rural areas). In the workplace it's far more common
69.
by smcleod 4mo ago
Cursor is one of the lower performing agentic coding tools these days. Seems weird to buy a VSCode fork.
70.
by smcleod 4mo ago
Those dense models are pretty fast with MTP now. 40-70TK/s depending on your machine, that's faster than cloud models (although not as smart obviously).
71.
by smcleod 4mo ago
Use the 27b, it's better in every way once you add MTP (which speeds up dense models but often doesn't add any performance to MoE models like the 35b-a3b). I get around 100TK/s on my 2x 3090 machine and 85 on my M5 Max.
72.
by smcleod 4mo ago
Your experience pretty much mirrors my own. I hate to be the 'they're holding it wrong' guy but there's certainly a lot of people out there that have no real idea how to effectively leverage AI.
73.
by smcleod 4mo ago
Hello, is the source available for inspecting somewhere? It looks interesting but I wouldn't trust running something with this level of access that I couldn't inspect the source of / run myself.
74.
by smcleod 4mo ago
Qwen 3.6 is far ahead of Gemma for most (but not all) things. I've deployed it out across a number of M5 MacBooks and it's genuinely useful for many tasks. It won't replace an Opus or current gen Sonnet sized model but it&#x2
75.
by smcleod 4mo ago
I don't see the point in comparing yourself to Haiku which is not only useless for coding but also old. No thanks Microsoft.
76.
by smcleod 4mo ago
Pretty low bandwidth and given how terrible nvidias software has been with the DXG et el, I would not put much faith into this.
77.
by smcleod 4mo ago
They release non-plus models to HF.
78.
by smcleod 5mo ago
That's impressive getting a 397B down to <110GB~. HF link is broken though!
79.
by smcleod 5mo ago
Really right now it's the M5 Max MacBook Pro 128GB, the RTX6000 is a nice card but you'd need more than one of them and you have to have a desktop to suit. The DGX Spark is slow and has pretty limited software support.
80.
by smcleod 5mo ago
Yeah that 60-150b~ range is such a sweet spot for current 'prosumer' hardware, I'd love to see something like a 120b-a14b or there about.
81.
by smcleod 5mo ago
Gemini is absolute garbage for anything useful, the last good model they released was 2.5 pro.
82.
by smcleod 5mo ago
I feel like it embodies Google's vibe of an uncool guy trying to stay relevant to the youth pretty well.
83.
by smcleod 5mo ago
How does it compare to context-mode or serina that are both well established now?
84.
by smcleod 5mo ago
That's the flash version not the full model and only at Q2-3~ so while impressive it's still quite different from the full model.
85.
by smcleod 5mo ago
This is the way it works in Australia. I wasn't a huge fan of having to prove who I am and have my identity related to be able to txt and make phone calls or use mobile internet when I moved here.
86.
by smcleod 5mo ago
The studio is really old now. The new one will drop at some point no doubt with more memory options. the 128GB M5 max MBP is great though
87.
by smcleod 5mo ago
Not surprised given how terrible Kiro is.
88.
by smcleod 5mo ago
TUIs reclaimed popularity long before Claude Code existed. Just look at the Charm ecosystem https://charm.land
89.
by smcleod 5mo ago
Apple Silicon before the M4 does not have matmul instructions which causes the prompt processing to be very slow. It's quite different on the M5, much like using a nvidia GPU
90.
by smcleod 5mo ago
Looks like it's https://github.com/systalyze/utilyze

More ›