I If I had to choose an online community that resonates with me, it would be Hacker News. For years, it's served as my muse, therapist, book club, and intellectual playground; all rolled into one. I deeply value the culture it fosters, especially the emphasis on thoughtful discussion.
Paul Graham’s essay "How to Disagree" remains essential reading for anyone engaging in online discourse . It provides a clear framework for constructive debate, and I agree that posts falling into the lowest forms of disagreement (ad hominem attacks or name calling) deserve to be flagged.
Yet, I share your concern, sometimes a post isn’t inherently bad, but attracts low quality replies. Flagging the entire thread in such cases feels disproportionate like amputating a limb just because there’s an itch you can’t scratch. It risks silencing potentially valuable discussion due to the behavior of a few.
I empathize with the moderators. Their job is thankless and difficult, and I appreciate that the warnings we see aren’t automated bots but messages from real humans trying their best. We all have limits and that’s ok.
Are all these improvement over custom kernel efficiency code ? Can we bring these to consumer RTX and Pro cards ?
After I read the article :)
The improvements in FuriosaAI's NXT RNGD Server are primarily driven by hardware innovations, not software or code changes.
This is so fascinating. I have been snowboarding for at least 15 years and had no idea that this was done by hand. The maps are so clear, specifically when you have multiple mountains with many one-way slopes across them. Next time I mindlessly throw away these art pieces, I will remember “Monet of the mountains”. Also will definitely share this story with someone on the chair lift ride.
I think the author slips into the same pattern he’s criticizing. He says LLM fans shouldn’t label skeptics as “afraid” then he turns around and labels the fans as “insecure” or “not very good at programming.”
It’s the same move; guessing what’s going on in someone’s head instead of sticking to what actually happened and what the tools can or can’t do.
The simpler truth is LLMs are great in some cases and painful in others. They shine on boilerplate and tests. They struggle when the domain is unusual, requirements are fuzzy; mistakes are made, you pay a big babysitting tax.
Instead of psychoanalyzing each other, people should share concrete examples
I totally agree the comment threads are definitely wild. People always assume we are against the technology. We expect more precision from machines, and that’s why they exist. Many people getting killed in car accidents is irrelevant to what we saw on the video. We expect more from them and their creators, and we do not want to see experiments on roads where we drive our kids.
I tried placing trackball on different positions especially with ploopy nano trackball and then with readymade options like UHK; none of them were comfortable enough. I now use the Logitech trackball which feels very convenient.
Random thought: Why don’t we have a server rack that runs at home but is managed by a company, then people or businesses rent them? Th important use case is that the heat generated can be used for home. Like property managers, business maintaining them are heat managers for home.
I feel like this idea would work better at a somewhat larger scale, like a small to medium datacenter heating an apartment or office building. The downside is for any of these systems is that when it's too hot outside that heat becomes a liability so you'd have to have the infrastructure to divert heat as well. The other downside is that you'd be replacing a very well understood technology with minimal maintenance requirements with a relatively complex technology with more extensive and complex maintenance requirements.
All the problems you mentioned can be reframed as positive feedback for the economy we’re evolving into. Let me dream the arguments for each of your bullet points:
1. Power fail-over (battery + generator backup) in every house?
- I recently listened to a Planet Money episode about how DC/AI infrastructure needs are driving up electricity prices in Ohio. Ordinary households end up paying higher bills while big entities plan/build for reliable power.
- Maybe household-level infrastructure could be improved as part of making this kind of model viable. This applies to networking infrastructure too
3. Could get expensive flying a technician to every household to upgrade hardware in the racks
- People with enough education can be trained, and with the incentive of being paid, households themselves could become the technicians.
4. Probably don’t want everyone at home having physical access to storage devices
- Same idea: if households are being paid and it’s “their role” to manage, the access concern gets reframed as operational responsibility.
5. Massive theft risk
- Theft risk already exists today (even in good neighborhoods). The incremental risk might be negligible.
6. Homeowner’s insurance would probably…
- If we squint hard enough, there are arguments here too (e.g., payments not missed, additional compensation).
I don't have the full data on hand, but there are three companies in the UK working on this problem. Feel free to go down that rabbit hole.
from google AI, YSRMV (Your Search Results May Vary).
Heata: Attaches servers to your hot water tank, providing free hot water during trials, saving users money on bills.
Thermify (SHIELD Project): Installs "HeatHubs" (often with Raspberry Pis) in sheds or homes, running remote data centers and using the heat for homes, aiming for low-cost heating for tenants.
Carno: Offers devices like the QB1 (digital boiler) and QH1 (computer heater) that integrate microprocessors into heating systems, reimbursing electricity costs.
The value doesn't line up for anyone. From my perspective as a homeowner, gas heating is cheap, and replacing my gas furnace with someone else's server rack is of significantly more negative value than my current cost of heating.
I don’t heat my office, the reason being, when I run ablations for personal LLM research, it is enough for heating my office during winter. Given our move from just searching for links or website content to well-summarized content, maybe enthusiasts can have a private inference cloud where GPU sitting idle can be used for heating and generate cents with inference. But yes, all those points are valid, and that could be a moat for the business.
I had a similar random thought: Why don't we have a cloud company that does something opposite to what a Cloud company usually does. Instead of renting out resources from a data center, they help company setting up local resources and sell/rent out the unused capacity to other companies/individuals?
Probably a bit of a nightmare to oversight all the resources and ensure consistency and privacy but hey why can't we dream?
If you put a rack in a datacenter, and it's a decent datacenter, you'll probably have at least 2x10G connectivity off the rack; if you need more, you can get more. If you put it at someone's home, good luck.
Conductor of an orchestra and current skillet of managing a coding agent feels intuitive to me which addresses understanding the differences in management and skill set. Great analogy.
Paul Graham’s essay "How to Disagree" remains essential reading for anyone engaging in online discourse . It provides a clear framework for constructive debate, and I agree that posts falling into the lowest forms of disagreement (ad hominem attacks or name calling) deserve to be flagged.
Yet, I share your concern, sometimes a post isn’t inherently bad, but attracts low quality replies. Flagging the entire thread in such cases feels disproportionate like amputating a limb just because there’s an itch you can’t scratch. It risks silencing potentially valuable discussion due to the behavior of a few.
I empathize with the moderators. Their job is thankless and difficult, and I appreciate that the warnings we see aren’t automated bots but messages from real humans trying their best. We all have limits and that’s ok.
reply