Just sharing my $0.02 here - I have ethical objections to using OpenAI or Anthropic products so I was a reluctant adopter of LLMs at all. Local models address most, though not all, my moral objections so I’ve been using them for work and personal projects for about a month.
The hardware I have (32gb Macs and a gaming PC with 10gb 3080) can only get me to Qwen3.6-35B-A3B at various quants but that’s enough (200-400 PP, 20-30 TG).
It’s taken some time to learn how to best utilize it - some things take a bit of babysitting or direction - but it’s quite useful. Not having ever used CC I can’t compare but it’s been a great assistant or pair programmer for everything from embedded C++ to Vue. I wish I could run 27B as there have been moments when this model feels like it just can’t quite figure something out but those moments are quite rare. For a lot of tasks it’s a huge time saver and has proved super capable at digging into and fixing bugs given pretty vague instructions.
I’m making a digital ESP32 powered synthesizer at https://subalpinecircuits.com/. It’s been a ton of fun, learning so much about every part of the process (and I don’t use LLMs). Currently I’m learning FreeCAD and figuring out what my case design will be. Woodworking, CAM and DFM is a whole other world for a software guy!
This nails my primary frustration with all gen AI - why are we all seemingly okay with a few massive companies and their billionaire CEOs training models on the output of all human civilization and then selling it back to us with the promise of putting all workers out of a job? How’s that not theft?
The electricity aspect has always been one I wish AI skeptics/critics wouldn’t lean on so much, because it seems like the one most likely to be solveable. I think we should be focusing on the economic aspects instead - if all this pans out as boosters promise, how does our economy exist when all white collar jobs go away? Who buys all the stuff, with what money?
I discovered the same thing with u8g2, and digging through the abstraction layers it felt like improving it was going to be impossible. Sending a single transaction with a framebuffer is so much simpler and faster.
SSD1322 looks great and might be something I look at for the future..
Interesting! That could be good way to boost the speeds here for sure, as I'm still pushing out a full framebuffer out with every update and am not usually updating the whole screen.
(author here) I've been pondering this, yeah. I'm currently sharing the I2C bus with a DAC and that's working alright, but the refresh rate issue is enough to make me consider SPI. I know the SPI peripheral supports DMA as well, and the I2C one doesn't (sort of? I know there's "async" transmit now but can't tell if that's really doing DMA)
I find this so confusing - I’m working on a fairly large project using ESP-IDF and find it an absolute joy to use. Far better than any other microcontroller framework.
> Abusive and spammy behavior. When abuse or manipulation of our service is
reported or detected, we may take action to limit the reach of a person's Tweets. Learn more about actions we take, including temporary and permanent account
suspensions, and limiting account functionality.
> Oh joy, another paper for the antivaxxers to misinterpret. Some red flags though.
> The tone of this paper, particularly the last sentence ("potentially harmful effects") seems suspect...almost as if they were intentionally riling up the anti-vax crowd.
> They appear to have misinterpreted the outcome of this study, claiming "higher risk of infection than unvaccinated individuals 9 months post-vaccination". The study makes no such claim that I can find. The tables in the supplement show the only subcohorts with a slightly higher incidence rate for vaccinated individuals at 9 months is age >80 (1.3 per 100,000 person days in vaxxed vs. 1.1 unvaxxed, supplemental table 5) and those requiring in-home care (1.6 vs. 1.5). Also notable is that the only vaccine that shows less effectiveness against infection is the non-MRNA vaccine (supplemental table 6) and that vaccinated individuals in all subcohorts were still substantially protected against hospitalization or death compared to unvaxxed.
> The primary source for the claim that mRNA-LNP is inflammatory is a paper authored by three of the four authors of this paper.
> Also notable is that at least one of the primary authors - the one credited with writing and editing - is a GBD-adjacent, anti-MRNA, HCQ truther (Don’t Look Up” Your Science—Herd Immunity or Herd Mentality? - "The virus fatality rate did not support lockdowns, blanket restrictive measures, and non-selective mass vaccinations. There was no solid scientific rationale to adopt the untested mRNA-LNP platform over other well-established vaccine formulations to fight COVID.").
> None of this means the science in the paper is wrong, but I'd be hesitant to take anything at face value.
Thank you for TenFourFox! As recently as last year I was still regression testing my code against TFF on a 12” PowerBook because it’s still my favourite hardware. If anyone ever used Bandcamp on a PPC Mac, they also have you to thank :)
Discover on the Bandcamp homepage is good for this, and we have pretty good genre-specific tag pages now that have a combination of algorithmic and curated picks, and you can dig in real deep with your own filters if you want. Example: https://bandcamp.com/tag/electronic
I work on the discovery team, so this is great feedback to have, thank you :)
All of that is fantastic! And as a relative nobody independent artist myself I totally agree with you, it's why I work there. Thanks for using the site!
I can't comment on our plans, but I personally totally agree with you and would love an API (and will continue to advocate for one!). Supportify is super cool :)
This is cool.. I was trying to figure out how to get macOS next/prev controls to trigger events in Safari but haven't had any luck. This seems like a cool API for doing that sort of thing (though, sadly, no Safari support)
The hardware I have (32gb Macs and a gaming PC with 10gb 3080) can only get me to Qwen3.6-35B-A3B at various quants but that’s enough (200-400 PP, 20-30 TG).
It’s taken some time to learn how to best utilize it - some things take a bit of babysitting or direction - but it’s quite useful. Not having ever used CC I can’t compare but it’s been a great assistant or pair programmer for everything from embedded C++ to Vue. I wish I could run 27B as there have been moments when this model feels like it just can’t quite figure something out but those moments are quite rare. For a lot of tasks it’s a huge time saver and has proved super capable at digging into and fixing bugs given pretty vague instructions.
I’m using Pi as my harness.