I'm building a temperature controlled dough proofing enclosure based on ESP32, an IKEA Samla storage box and an inexpensive silicone rubber heater. It's delightfully impractical-- purely a first hardware/MCU learning project and a holiday gift for a relative who I think will appreciate it.
I'm impressed by how far I can get "vibe making". Most of my professional experience is in high-level software, but AI gets me unstuck quickly when I don't know something specific to ESP-IDF or the hardware. As of today I've got a circuit tested, firmware nearly complete, and a custom PCB en route from JLCPCB.
One limitation I’ve noticed: ChatGPT struggles with the details of part selection (e.g. choosing specific temp/humidity sensors or connectors). Adding datasheets to the context helps a lot, which makes me wonder why this isn’t something the model can do or at least ask for.
This is nonsense and your cynicism is unwarranted. I personally know Scott. He's worked on food tech in Seattle for decades, and developed the product and fundraised for Seattle Ultrasonics locally.
Ultrasonic knives are historically large footprint devices used in commercial/industrial food prep. The innovation here is making ultrasonic hardware compact enough to fit in the knife handle.
I was also pleasantly surprised by the quality of this article, especially the nuance around the value of a growth mindset. It's consistent with my experience of having had to work pretty hard to learn programming.
There's plenty of greenwashing in tech, but I don't think it's hopeless. On an ops call this morning I heard about a team switching to a server type that saves ~40% on power, and their fleet is ~50k instances. Opportunities exist if you have the skills and this is an issue you care about.
But the net effect is usually not less is it? Management will find a new way to spend that money or the team will realize they can now run 40% more instances on the same budget.
I'm a customer and fan of interviewing.io but I'm also disappointed in this post. It's not actionable information for candidates.
A lot of the interview prep industry deliberately stokes interview anxiety to drive sales. I appreciate that you folks have a product (and marketing tone, usually) that emphasizes results over hype, but I think this post falls short.
Have you had discussions about this with your manager, and made it clear you're interested in working on different things? How did it go? You probably have more agency in this situation than you think.
I did light research into a similar idea, and got stuck linking aircraft registration numbers from ADS-B to businesses. Many aircraft owned by corporations and high net worth individuals are registered to subsidiaries or holding companies. I wasn't able to follow the chain of registrations through opencorporates.com.
I probably didn't exhaust all possibilities -- wonder if this is a solved problem.
I've found The Five Dysfunctions of a Team useful. It's a clear description of things that go wrong when people work together, even when everyone has good intentions.
I'll second the suggestion to speak with a professional therapist or coach if it's an option. These situations are difficult to navigate and a neutral perspective can be helpful.
If you'd consider an (inexpensive) paid option, https://www.ckbk.com/ takes an interesting approach. They partner with publishers to put cookbooks online. Fairly powerful search/filters, and the focus on cookbook content is a good proxy for recipe quality.
I'm not affiliated with them, but Matt (cofounder) has a background in scientific publishing and is passionate about recipe sharing.
Are they verified recipes? I’ve found that a lot of books just stole recipe and never tested similar to how online does it now. Also lots of recipes are needlessly complex
I'm impressed by how far I can get "vibe making". Most of my professional experience is in high-level software, but AI gets me unstuck quickly when I don't know something specific to ESP-IDF or the hardware. As of today I've got a circuit tested, firmware nearly complete, and a custom PCB en route from JLCPCB.
One limitation I’ve noticed: ChatGPT struggles with the details of part selection (e.g. choosing specific temp/humidity sensors or connectors). Adding datasheets to the context helps a lot, which makes me wonder why this isn’t something the model can do or at least ask for.