I find this to put my psql in a better place as a daily driver. Adding this to ~/.inputrc [1] will allow you to limit search on your entered text. Eg.
select (up arrow)
will loop through your psql history for commands that started with e.g., . The challenging part is in wide tables and or table with large data. Less is awkward usually so using pspg made it less awkward.
I tried also to with help of ai, to write a plugin for sublime that fits my flow. It worked well but I think I'm more used to psql.
in my region (middle east) facebook marketplace is the defacto listing. I know craiglist (just from me being on internet for so long) but i don't think it operates here. And for sellers they also only know facebook.
I'd rather not post invite links here directly out of concern for spamming, but I can leave some easy-to-follow breadcrumbs.
The decomp.me Discord server invite link can be readily found on that website and in the README of its GitHub repository. It's the closest thing to a central hub of the decompilation community. You can find some invite links in its chat history by searching for "discord.gg" (including the servers listed below).
Some of these Discord servers have a #other-servers or #related-servers channel with tons of invite links to other similar Discord servers. In particular, these servers have those channels:
- PS1/PS2 Decompilation
- GC/Wii Decompilation
That way, you should be able to find dozens and dozens of Discord servers on that topic. There are still many more out there (I've joined at least six others that aren't directly reachable from the invite links inside the servers I've mentioned).
There are also other Discord servers about reverse-engineering that can contain discussions about decompilation techniques or projects.
I mostly don't use Discord, but I'm intrigued how this system works. If the invite link is already posted on the project website and the readme of a Github repository, why would putting it in a Hacker News comment risk spammers?
First, because personally I think it's bad etiquette to post an invite link publicly on large, open-access forums like Hacker News.
Second, because while the biggest decompilation Discord servers have effective moderation team and processes, the vast majority are just a server for a project from one or a couple of persons with <100 members joined. Such small servers don't have round-the-clock moderation or customized settings.
Third, because I regularly see phishing crypto spam posted in the smaller servers. Spreading invite links publicly carelessly increases the odds that these scammers find them and spam them.
The part I don't understand is how posting the link on HN is different from posting it on Github. I'm wondering for example if Github has some sort of extra bot protection HN does not, or if there's some cultural difference I'm not aware of or seeing.
Discord servers are more like a hangout place than a public forum and public invite links are like putting a key under a mat. Just because the owners of a Discord server put it on their public webpage or GitHub repository doesn't mean it's cool to spread it around willy-nilly.
There are also practical concerns. An invite link that needs to be deleted (for example due to spam abuse) means that it will no longer work. Updating all the places where that old link appeared to the new one can be impractical or impossible.
HN isn't a good place to share Discord invite links. I'd recommend putting an email address (or something) in your bio – ideally one you don't mind getting spammed – so people can send such things to you semi-privately.
The gemini/ai part aside, but I really like this revival of passion for PCs and Laptops. Totally anecdotal here and I could have definitely spend couple of minutes to research the marketing numbers, but I cannot help but feel happy with Framework, Panther Lake and Dell XPS and of course the mini Mac and Macbook family. I feel like there were years when center of attention had turned to mobile/ipads (and consoles) which were severly locked down to the point of no use point their intended creators purpose. I felt bad my siblings never get hooked into my old PC, as they went from PS3 to phones.
Previous cycles were fueled by retail, with an industry trying to legitimize itself.
This cycle is about max extraction and fraud - Legitimized by the presidential family cashing out billions in meme coins, insider trading and forks of existing protocols.
Hacks have also been hitting hard. North Korea has stolen 500m this year alone and 2b last year.
So… no thriving. On the opposite. Dying is a more appropriate word at this time. Some would call this an opportunity. I see more pain ahead.
No wonder Coinbase is laying off people with the excuse of AI. The reality is that volume is zero. At this stage only me and a bunch of other retail weirdos keep on buying bitcoin paycheck by paycheck…
IMO there's actually an opposite effect going on: Retail was never the main driver behind crypto pricing. Think of it like this: If every US adult purchased $1000 in BTC, that purchase would represent a level of volume that BTC, otherwise & at today's historically low volume, takes about 5 days to clear. BTC volume is too high to be explained by retail.
Crypto volume comes from institutional liquidity, not retail. All of that liquidity has moved from crypto to AI. It turns out that the liquidity wasn't actually interested in the technology or the philosophy; they were interested in outsized ROI. Think of BTC not as a currency, but as a share of stock in the crypto technology sector.
Idiots who were getting fleeced by shitcoin pump and dumps are now getting fleeced by insiders betting on heads of state being assassinated.
That's the problem with building your castle on a quicksand whose fundamentals aren't in the same order of magnitude as the market cap you command. When all you truly offer is gambling, eventually a shinier casino will open up and eat your lunch.
I tried it on a non trivial, but also well documented and self contained task. It did amazingly well. I used deepseek v4 pro via deepseek platform. The model is very fast and also it is super cheap. I burned only 0.06 USD (I reckon how the same task would have cost me had I used e.g., amp).
PS. mentioning amp because i used to use it and I pay directly for token. I topped up 5 usd so I will be going to use it and see how far can it take me. But my impression so far is even when model subsidization is done, those open source models are quite viable alternatives.
> But my impression so far is even when model subsidization is done, those open source models are quite viable alternatives.
My understanding is that DeepSeek V4 Pro is going to be uniquely good at working on consumer platforms with SSD offload, due to its extremely lean KV cache. Even if you only have a slow consumer platform, you should be able to just let it grind on a huge batch of tasks in parallel entirely unattended, and wake up later to a finished job.
AIUI, people are even experimenting with offloading the KV cache itself to storage, which may unlock this batching capability even beyond physical RAM limits as contexts grow. (This used to be considered a bad idea with bulky KV caches, due to concerns about wearout and performance, but the much leaner KV cache of DeepSeek V4 changes the picture quite radically.)
Good. It's hard to overstate how nervous most executives are about relying on cloud-based providers.
AI currently works basically by sending your entire codebase and workflow, and internal communication over the internet to some third party provider, and your only protection is some legal document say they pinky promise they won't train on your data.
And said promise is made by people whose entire business model relies on being able to slurp up all the licensed content on the internet and ignore said licensing, on the defense of being too big to fail.
Yes, this is the most straightforward argument for local AI inference. "Why buy cloud-based SOTA AI? We have SOTA AI at home." It's great that DeepSeek may now be about to make this possible, once the support in local inference frameworks is up to the task.
Is there any place I can read about KV? Excuse my ignorance as I'm not familiar with this topic and I read scattered notes that deepseek's cost are well optimized due to how their kv cache work. But I want to read more how kv cache relates to the inference stack and where does it actually sit.
> AIUI, people are even experimenting with offloading the KV cache itself to storage, which may unlock this batching capability even beyond physical RAM limits as contexts grow.
Especially this point. Any reason that this idea was considered bad? Is it due to the speed difference between the GPU VRAM to the RAM?
KV cache generally grows linearly with your current context; it gets filled-in with your prompts during prompt processing, and newly created context gets tacked on during token generation. LLM inference uses it to semantically relate the currently-processed token to its pre-existing context.
> Any reason that this idea was considered bad?
Because the KV cache was too big, even for a small context. This is still an issue with open models other than DeepSeek V4, though to a somewhat smaller extent than used to be the case. But the tiny KV of DeepSeek V4 is genuinely new.
> even when model subsidization is done, those open source models are quite viable alternatives.
Model inference was never subsidized. Inference is highly profitable with today's prices. That's why you have many inference providers. My guess, the prices for inference will go down, as more competition starts cutting the margin.
It's model training, development and R&D that cost a lot, and companies creating closed models don't have any business model except astroturfing and trying to recover training costs through overpriced inference.
>people are just using the latest and most expensive models because they can,
While I agree with the sentiment, I think that might have been initially driven by older models being nerfed and/or newer ones were better at token/$. And there is this notion that those labs don't constraint the model on the first days after its release.
I (codex) made a plugin for stremio to stream my collection from real-debrid.
I tried existing plug-ins first and non was working. Just prompted chatgpt to refine my initial specs, and asked on another session to build that. And later used codex for the last mile. Nothing fancy though and nothing can be particularly useful to others but damn it was too useful to me and my wife.
Ultrathink isn’t “removed.” Its behavior is different. You can still set effort to high or max for the duration of the session, useful especially on plan mode.
select (up arrow)
will loop through your psql history for commands that started with e.g., . The challenging part is in wide tables and or table with large data. Less is awkward usually so using pspg made it less awkward.
I tried also to with help of ai, to write a plugin for sublime that fits my flow. It worked well but I think I'm more used to psql.
[1]
~/.inputrc
$if psql "\e[A": history-search-backward "\e[B": history-search-forward $endif
edit: formatting
reply