scosman

They can plan, but no guarantee it will produce what you want. Sometimes most of the work is aligning on what to build. And I'm not handing over technical planning to it yet.

I use this skill and it makes the specing process progressive. Human driven for the "what", 50/50 for higher level technical planning, only where it has questions in the low level details: https://github.com/scosman/vibe-crafting

Oh, Canada wouldn't actually welcome a local monarch. Ceremonial across a sea... sure, why not. A King moving to our shores, nope.

I've done this exact mode on my Mitsubishi ACs. Despite being old units that came with the house, they are instantly responsive and report back current temperature flawlessly.

ESPHome device are consistently faster and more reliable than any smart home tech I buy at 10x the price (not hard with $3 boards).

I made https://froogle.fyi out of a desire to get that old school search back. Source: https://github.com/scosman/froogle

It hasn’t replaced Kagi as my daily driver but the UX is fun.

highly recommend clicking "Listen to this blog post". Not what I was expecting...

one box of cables?

those are rookie numbers

why is it jumping through hoops to get online if it's already online?

> “nobody asked for this”

I absolutely want a general purpose assistant.

I absolutely don't trust Facebook with the necessary data.

Ah yes. Using that OTA update it got via AM radio.

Reminder: you can connect it to your network, but block it from the internet at the router. That way you can still use Airplay/Chromecast.

I have an LG. Great OLED panel. It's never seen the internet.

Proxmox + community-scripts.org is pretty great.

I wanted to say "You actually can't fly a helicopter to the top of Everest, the air is too thin", but apparently it's been done exactly twice. Stripped down specially chopper and ideal weather conditions.

Community effort happening here to build the ideal dataset: https://github.com/scosman/pelicans_riding_bicycles

The year is 2031, the Meta LlamaBrush won't start until it's played a 30s unskipable ad for Doritos spearmint.

They were investing in ANE and Metal before everyone in consumer. Hardly asleep. They just underestimated the market size, as pretty much everyone did.

Right now sweet spot is voice transcription. Meeting recording apps are genuinely better locally than in cloud. Can run on an M1 easily. Latency matters. I built https://github.com/scosman/Biscotti and see zero reason to use cloud ever again.

LLMs are harder: not much useful below 12B, and the 700B+ ones are really much better. Models like Qwen 3.8 27b show promise: in a few years pretty good local AI should be in reach for anyone willing to buy a $1000 computer (but who knows what your $20 sub buys you then).

z.ai is using all Chinese hardware for flash: https://thenewstack.io/glm-5-3-flash-chinese-chips/

There are other providers with much faster inference, like BaseTen at >100t/s: https://openrouter.ai/z-ai/glm-5.3-flash#performance

You can't compare models released 6+ months apart. GLM 5.2 was same architecture as 5.3 and not nearly as good. Takes time to build frontier intelligence and distill down to smaller sizes.

It's actually slightly more expensive ($0.50 vs $0.48), but there's a temporary 50% discount.

I've seen dozens of conversations about it in last 24 hours, and every major inference provided added in first 24 hours. I think it's gaining plenty of traction.

z.ai coder plan, both in opencode and direct API access. I use it for my side projects like https://github.com/scosman/Biscotti (on-device meeting transcription and summaries).

I've been using it more and more. Feels like Opus 4.8, in the best possible way.

I just shipped this free app: https://github.com/scosman/Biscotti

Private offline transcription and summary. Speaker identification, working on voice-prints for identification across the corpus.

A new case where the "free as in speech, not free as in beer" clarification helps.

yeah, it's X that has ads. Which is why they wouldn't be as receptive as HN.

It's fast by computer standards and excellent for entry level chip. The Pro/Max/Ultra chips are always faster.

Compared to something like VRAM it's slow.

Note this is from tobi, Shopify CEO

Amazing he still codes. But likely more experiment than prod level.

it will still be on "private.icloud.com". Just as filterable...

ah, that makes it feasible! Okay, glad it's not technical limit. They should fix the pricing...

And GLM is only 0.7T!

But these labs distill off the larger models. Both officially at the labs with the big ones, and unofficially. We need the giant models to get the smaller models.

I don’t think they will until they change the architecture.

They don’t have a prefix cache like other providers, or at least don’t have a discount in their billing structure. Each message charges for the whole context window. It’s wildly more expensive for long multi turn scenarios with lots of tool calls (coding). It’s better for short few turn tasks.

Edit: I don’t know if they actually have a proper cache. This could just be a billing artifact.