bme's recent activity
-
Comment on What programming/technical projects have you been working on? in ~comp
-
Comment on What can be done about Linux being slopified? in ~tech
bme Link ParentI'm going to be slightly mean and say that if you are happy across the board with AI output then it's not in factory mode (i.e. you are sweating the details in the loop) or you aren't reading the...I'm going to be slightly mean and say that if you are happy across the board with AI output then it's not in factory mode (i.e. you are sweating the details in the loop) or you aren't reading the output or you aren't very good at software and this can't tell that the outputs are bad.
That is to say I think we agree? :) I was mostly trying to offer a corroborating view. I do a mix of embedded and dist sys (we sell hardware + associated software services).
-
Comment on What can be done about Linux being slopified? in ~tech
bme Link ParentI think it depends on the degree of freedom you give it. I typically always watch what it is doing and review it really hard. I'm still winning and as a team we landed a PR deleting 10kloc this...I think it depends on the degree of freedom you give it. I typically always watch what it is doing and review it really hard. I'm still winning and as a team we landed a PR deleting 10kloc this week. AI doesn't have to inexorably lead to bloat. We have landed many changes each savings worth thousands a month on AWS just by doing simple things like pointing an agent at metrics for a service and configuration and catching low hanging fruit like "your compression and batch settings have drifted from reality". Stuff that sure, anyone could do if they had a spare 30m, but Claude can make an artifact in 2m + short MR, that can be canaried in production in less time that it would have taken a human to do.
Do they write good code? Eh. I keep hearing that you can skill your way into making the bot behave like you, I don't buy it. Or I am in the same boat as you and I suck at writing skills. I really am bullish on ai for being in the details, but not on writing excellent code yet.
I really like it for triage and assisted code writing. I don't like it at all in "software factory" mode. Anthropic can't even get scroll back working properly Linux in Claude code.
-
Comment on What programming/technical projects have you been working on? in ~comp
-
Comment on SpaceXAI joins Omacom and gives $1.5m Grok tokens for Omarchy in ~tech
bme LinkI wish we would all stop talking omarchy. It's a nothing project, bringing nothing new or interesting into the world. I had another exec the other day ask me if I was going to try and listed off a...I wish we would all stop talking omarchy. It's a nothing project, bringing nothing new or interesting into the world.
I had another exec the other day ask me if I was going to try and listed off a bunch of shit that it "has", that effectively was like yes, of all the distros, truly, this is one of them.
Not to mention I currently run NixOS, and moving off it to something less replicable would be just laughably stupid at this point.
-
Comment on Repo for syncing ipods devices on Linux in ~tech
bme Link ParentThere are a bunch of arguments being made here. I am only focusing on the quality one. It is completely obvious that you can improve the quality of your software with AI. I am running a team of 5...There are a bunch of arguments being made here. I am only focusing on the quality one. It is completely obvious that you can improve the quality of your software with AI. I am running a team of 5 people and merely by offloading initial triage of inbound customer support requests to AI we are fixing more bugs. No code gen required. Just let it rip on cross correlating logs and traces. Does it sometimes get it wrong? Yes. Is that easy to spot? Yes. Is the software getting better faster because of AI? Yes. You can also use AI for background security hardening and performance testing. In both cases you can have verifiers that don't depend on AI. It can run in the background while you do other stuff.
Lastly there are plenty of features that AI can one-shot faster than most software developers at equivalent quality or better. That might sting but I think a lot of people massively overestimate the talent of the average dev. It's not pretty out there.
Saying all AI usage in coding is slop and negligence and inevitably makes software worse is unserious and will redirect the debate away from questions that matter (IP theft, environment, deskilling, labour displacement with zero plan).
-
Comment on I'm having a glow up, but it sucks in ~life.men
bme Link ParentYou managed to get hot. Getting hot != easy. You can conquer bureaucracy. You are here typing in coherent and engaging English. I believe in you! Also what @Akir said.You managed to get hot. Getting hot != easy. You can conquer bureaucracy. You are here typing in coherent and engaging English. I believe in you!
Also what @Akir said.
-
Comment on A24, Sony and The New York Times circle Letterboxd as bidding starts at $300 million in ~movies
bme Link ParentI can see that argument. I think where it runs into difficulty is that, the owner of a business can cease trading for any reason. Money doesn't need to change hands. Is that immoral? I don't mind...I can see that argument. I think where it runs into difficulty is that, the owner of a business can cease trading for any reason. Money doesn't need to change hands. Is that immoral? I don't mind the idea of trying to steer sales towards the better new owners and that money shouldn't have primacy in all things.
What if the only buyer is say a PE company, or some other company that some people might feel will make the service worse? What if the service as it stands today is unprofitable and it will require new management with the will to squeeze it more to keep it above water? Should the business just go under? Who is deciding which owners will do the right thing? To be able to brand a sale immoral requires a prediction of behaviour. I guess I'd rather than adjudicate on the character of the buyers I'd be more interested in trying to legislate or regulate the behaviour we would like from companies. We should have more b-corps.It also seems to me that dependency is also doing a lot of heavy lifting in this argument. Letterboxd for instance seems to me like a thing that doesn't need to exist. I don't think restrictions on its sale should be given much thought at all.
Anyway, if you find any of this tiresome or obnoxious, that's not my intent. Feel free to tell me you don't find it an interesting conversation :)
-
Comment on A24, Sony and The New York Times circle Letterboxd as bidding starts at $300 million in ~movies
bme Link ParentDo you think people that start companies should be blocked from selling them to willing buyers at a fair price? Aside Monopoly/Market consolidation reasons, I don't see how this is close to...Do you think people that start companies should be blocked from selling them to willing buyers at a fair price? Aside Monopoly/Market consolidation reasons, I don't see how this is close to reasonable.
More specifically: how is it morally wrong for someone to make something and sell it on to someone else, even if the new owner won't run it as well?
-
Comment on Googlebook revealed and ready for pre-order in ~tech
bme Link ParentSomehow I missed the bit about flying cars lolI am sure you are being a little flippant here but fling to device implies a knowledge in space of where other devices are. That's not easy. Unless you have some kind of always on ble mesh that everyone participates in to do network tomography and generate points in space and the devices are dense enough and diverse enough to make the derived coordinates accurate.Somehow I missed the bit about flying cars lol
-
Comment on Where do we go from here? (Regarding AI) in ~tech
bme Link ParentI got my feet wet with this during the pandemic and at one point materially affected the global supply of raspberry pi 3bs! You are correct that supply chain agility, and being able to respin a...I got my feet wet with this during the pandemic and at one point materially affected the global supply of raspberry pi 3bs!
You are correct that supply chain agility, and being able to respin a board is really important when you don't have someone else financing your supply chain 6m to year in advance. The problem you then land yourself in is maintaining firmware for all of your board variants. We have been shipping "universal" bundles covering off 6 different pic micro controllers for 5 years at this point, even though we don't even use those parts any more!
More recent example: we wanted initially to use kioxia emmc. Western contract manufacturer (with global footprint, hundreds of millions in revenue) in 2025: we can maybe get you some towards the end of 2026. Move to manufacturing in the far east and all of a sudden they are just available. Now of course they are available at a price we don't want to pay, so we are looking at drastically shrinking storage to 32mib of external spi nor flash. Argh!
No blog. I'd dox myself so fast.
I wish you all the best with your endeavours, sounds cool!
-
Comment on Billing for web hosting in ~comp
bme Link ParentThe thing you want to care about more than uptime monitoring imo is payment method monitoring. My greatest shame (and I think about it all the time), was I set up a website for a friend and my...The thing you want to care about more than uptime monitoring imo is payment method monitoring. My greatest shame (and I think about it all the time), was I set up a website for a friend and my card expired and I missed the email and the whole lot got nuked.
Hosting someone's site is a big deal. The tech is easy but they are entrusting you with the data, and when it's gone it's gone. How good are your backups, can you restore them blah blah.
-
Comment on Where do we go from here? (Regarding AI) in ~tech
bme Link ParentThe supply chain has also gotten really weird. I don't procure directly but I am involved with parts selection for the bill of materials for let's say mid-tech consumer electronics. Lots of things...The supply chain has also gotten really weird. I don't procure directly but I am involved with parts selection for the bill of materials for let's say mid-tech consumer electronics. Lots of things can't be bought even though the initial evidence is telling you that you can. The buying is effectively monopolised among the world's largest contract manufacturers.
So even if you have all the will, unless you can fab everything someone else will always have a say in what you can get. Even commodity parts like emmc or non-cutting edge stuff like 30nm chips, low memory ics like 64mb Linux-capable SIPs.
-
Comment on It breaks a village in ~comp
bme Link ParentI think the problem roughly is that AI is a multiplier, and plenty of people probably need a divider. I don't think it's additive. The damage of someone bad was limited. The limits are off now (at...I think the problem roughly is that AI is a multiplier, and plenty of people probably need a divider. I don't think it's additive. The damage of someone bad was limited. The limits are off now (at least until companies start getting charged something somewhat connected to the real costs).
I am getting far more done with far more thorough testing than previously, and I've almost got to the point where I've gotten ai generated comments to manageable.
When I hear about people talking about total collapse of projects due to ai, I mostly think they are speed running the fate they were likely to hit eventually.
-
Comment on What are your current pain points with Linux? in ~tech
bme Link ParentThis is basically where I am. My Linux problems are externally imposed, mostly by hardware, mostly sleep/suspend.This is basically where I am. My Linux problems are externally imposed, mostly by hardware, mostly sleep/suspend.
-
Comment on What are your current pain points with Linux? in ~tech
bme Link ParentIncredible comment sent from my NixOS laptopIncredible comment
sent from my NixOS laptop
-
Comment on Your executable is a SQLite database in ~comp
bme LinkRidiculous work. So good. Honestly love it, and very on brand. I would absolutely love to see a persistent sqlite process, write lock held by a nix Daemon, readers able to enjoy lighting fast...Ridiculous work. So good. Honestly love it, and very on brand. I would absolutely love to see a persistent sqlite process, write lock held by a nix Daemon, readers able to enjoy lighting fast access to the database. It does seem like not being able to share memory is lame but if statically linked distros exist then it's at least practical if not efficient.
-
Comment on Elixir Is All You Need in ~comp
bme Link ParentThe beam ecosystem is nice. I recently wrote a small sidecar to do something very boring but quite challenging to do well, and chose elixir: OTP Libcluster Nicer language than Erlang Reasonable...The beam ecosystem is nice. I recently wrote a small sidecar to do something very boring but quite challenging to do well, and chose elixir:
- OTP
- Libcluster
- Nicer language than Erlang
- Reasonable memory footprint
All of this ended up being true, but I did not leave a fan of elixir. I'd say decidedly mid as a language. Not sure what the fuss is. I think gleam looks much more pleasant, but it currently enjoys the advantage that I didn't have to do anything with it.
The thing I had to do: we run at work a couple of stateful HA systems that can tolerate loss of VM and will re-replicate the lost state. However, this is expensive in time and bandwidth. Better if you separate compute from storage (we are on Aws). So I wrote a small sidecar that scans for ebs volumes and attaches them if they exist such that a fresh VM can assume the identity of a departed VM. This gives you a stateful set on steroids. Libcluster gives you a cluster with no effort, erpc gives you instant native rpc, OTP gives you a great framework for resiliency. I am sure go + go-discover + serf probably gives you a similar-ish base, not actor mode baked in, and for a collection of simple stateful endpoints actors are perfect.
-
Comment on I'm starting a small business: any advice? in ~finance
bme LinkGeneral advice: winning mostly means not losing for long enough. Avoid over extending. Almost everyone I know that has lost their shirt did it because they over-extended on a deal that went sour....General advice: winning mostly means not losing for long enough. Avoid over extending. Almost everyone I know that has lost their shirt did it because they over-extended on a deal that went sour. You can be 100% right and still get squished if you can't ride it out.
-
Comment on Honey gets terminated as lawsuits proceed in ~tech
bme Link ParentIt's not (just) that. It's that they charge the retailer a bunch of fees (significantly above regular card fees). Which of course is eventually reflected in the price if enough people do it.It's not (just) that. It's that they charge the retailer a bunch of fees (significantly above regular card fees). Which of course is eventually reflected in the price if enough people do it.
That makes sense. I've always felt like this is the end game problem to solve. Just imposing a speed limit seems like a good start. I've always liked the idea of self diluting authority, but simplistic models (don't capture dynamics over time).
Cool project!