[ my public key: https://keybase.io/whs; my proof: https://keybase.io/whs/sigs/XsIzvhNTTXrZgjL36BBv_SxD8KYmsGihbLIaFJ5WAB0 ]
openpgp4fpr:114f08a7e2081aca01ec7738461ccb345ad1e4a5
Submissions
Google Cloud starts charging for public IPv4 on January first, 2020
The only two reasons it was worth it is my company paid for it, and you can get a lounge space and some swag at reinvent. I never get to use it professionally
I've been looking for similar distributed computing style LLM, and I found AI Horde and a few other smaller efforts like one from Aphrodite people and distributed training from Nous Research.
AI Horde seems to be the biggest of them all. Their API speaks KoboldCPP text completion (not even chat completion). It seems that the community (or at least the active people) strongly prefer it this way because the API exposes more tunables than chat completions, which for roleplay use seems to result in better result. I don't know what else you can use AI Horde for anyway since all other use cases likely will require tool use. Just this week I was set out to improve their OpenAI bridge to support chat templates and response parsing. We'll see if I could get it deployed officially then you might be able to use it to code, although you'll have to use RP models.
I think Horde do have a lot more abuse prevention. Workers needs to have 1 week of cumulative uptime to be considered trusted to prevent brigading - users can opt into trusted workers only. Running a worker give you kudos which is required for >512 max tokens generations and also free requests get bumped to last.
I wonder how security is done in this engine, since it's accepting input from anyone. llama.cpp's RPC layer seems to says that you shouldn't run it in public (I assume because it is lower level and may result in RCE on your GPU)
Most people don't directly call Linux syscalls though but go through glibc. It might even be unavoidable if you want to ship desktop apps as the library will use it. If it's that easy there wouldn't be Python's manylinux, flatpak base packages or Steam Linux runtime
My company is heavy on Cursor and I still ask them to provide me GitHub Copilot, for the sole reason that Cursor is probably the reason Microsoft had to implement technical enforcement of their TOS on proprietary plugins. Previously, you could use PyLance on VSCodium but now those plugins do not work outside VSCode anymore.
If Cursor (and every other commercial VSCode forks) didn't use MS extension store in the beginning and violate the TOS these might not have happened.
I think AWS billing is quite complicated that they probably don't even know what did you get charged specifically for this machine.
You might have leftover reserve instance that applies, which make the listed price inaccurate. That reservation might even be in a different AWS account in the same organization that you don't have access to. That reservation might not even be there between the time you quote and the time you actually launch it if someone/something did launch before you.
Your organization might also have discounts. I believe some discount may also be very confidential. For example, my reseller policy is that the customer must not be able to see AWS Billing in the organization root account as supposedly the price in that console are the price AWS charged the reseller, while we pay listed price minus any discount we negotiated ourselves.
Finally, I suppose they don't want to have prices shown in multiple places as they will need to update it when prices changes. Doesn't want to risk forgetting one place and getting sued for it. You can see that AWS documentations often do not want to mention the price at all, even if that price is currently free.
Chinese clouds kinda make this simple by making reservation part of the buying machine itself - you have to mark that particular machine as monthly/yearly committed when you start it (or convert it later). The complicated part is recycling instances - if you delete a server before its reservation ends it ends up in a recycle bin that you need to look before making new reservations.
I wish that is an option. Nowadays many non competitives games that you play with friends you trust still use EAC (yet accept non-kernel mode operation on Linux). I suppose other than VAC you can't buy a usermode anticheat middleware now.
I wrote gRPC xDS server for Kubernetes that is configuration-free. Basically just load xDS client library into your code, then use xds:///servicename.namespace (unlike DNS, namespace is always required). It should be as lightweight and scales in similar way to the cluster DNS.
My company run this exact code in production since it was created in 2022. We probably have several times more than 1000 rps gRPC requests running internally including over the public internet for hybrid cloud connectivity. That being said, gRPC's xDS client is not always bugs-free.
I think we ended up in this situation because of this outsourcing. Competitive games and MMO need comprehensive security solution, as cheating has a global lasting impact in matchmaking. Attackers may also have financial motivation to attack the anti cheat in these games.
Coop games might don't need as much security as competitive games, as some games do not have global state, or the global state is simply cosmetics. Since nowadays all the anticheat you can buy (except VAC) are kernel mode you'll have to accept the security risk just to have fun with your friends.
I tried it and contributed to searx. It didn't give the same result as Google, and it also have 10k request rate limit (per month I believe). More than that you'll have to "contact us"
It should tell me what should I look before I trust it. Not trusting the workspace means I might as well use Notepad to open it. I wouldn't think that tasks.json include autorun tasks in addition to build actions.
Other people have described Factorio-style game, or idle RPGs. I think for a Factorio-style game, Desynced might be a better programming game. I haven't play it for quite some time, but last I remember it has a visual editor and automation bots in early tech tree. In Factorio, if you craft a green circuit board it automatically craft all the precursors from iron/copper plates in your inventory. In Desynced you have to do all that manually, which force you to use the automation system right up front. However, it's not fun repeating code blocks to build different simple things, and it's not fun that you have to solve both gaming problems (eg. I need to build my base, what should I build next) and programming problems (eg. write flood fill bot to explore for resources) at the same time.
The problem with these kinds of game is at some point you'll run into tech debts - your factory layout is not optimal, and the game optimize your starting zone so migrating out means you may lose simple access to starting resources. I tried optimizing for large factory up front, but it's like a startup with a monolith running on a Kubernetes.
Satisfactory solved this by having a fixed map and 3D, but it comes with its own challenges. Good Company partially solved this by replacing belts with humans who can walk anywhere you want, and introduce belt at a late game stage (which I quit around that stage - the game already gets repetitive by that point).
Anyway, what I was describing was not those kind of automation games. RuneScape is an RPG game with over 20+ skills. What make it interesting is that RuneScape bot engines (which exists, but is illegal) do provide all the high level primitives for you in Java. You could provide a world coordinate and a walking script will do all the walk for you even if it is the other side of the world map. It should felt like making games in Scratch instead of reinventing serialization in Screeps.
Other games with scripting also often don't allow human-in-the-loop. If a bot in an automation game get stuck because you forgot to program how to restock teleports, you have to stop it and reprogram. RuneScape bots don't block inputs - if you forgot to make the bot handle stamina exhaustion just click the run button yourself without stopping the script.
RuneScape also comes with strategizing for the human, while the solution space is quite fixed and well discussed by the community if you don't want to find out yourselves. If your Slayer task is green dragons, do you go to a spot that is far from a bank and lose efficiency or give up the loot, or do you go to PvP zone and use cannon (multi target auto turret)? If I write a bot, I probably will write a Wildy Green Dragon script instead of a generic fight-anything script so that it will know to bring cannon and place it at the optimal spot to target all spawns at once. Same goes for many other skills - if I train crafting or construction do I sink millions and go for the highest XP rate or do I go very slowly and get a slight profit out of it.
I'd also add that having a community marketplace would also helps. RuneScape have underground marketplace for bots, with free scripts that probably get you banned, and paid "private" scripts that supposedly undetected. I also have played Mars First Logistics where you build a vehicle Lego-style, then deliver weird cargo like a block of ice or a crate of oranges without a lid. The game have Steam Workshop support where you can just skip building your own vehicle and try to drive someone else's vehicle to destination. I got an ice block pusher, which I need to figure out how to put the ice in (just surround the block, close gate so the block cannot escape), then the map has hills that make your ice fall over the bottom anyway.
>Google: Google does not offer a public search API. The only available path is an ad-syndication bundle with no changes to result presentation - the model Startpage uses. Ad syndication is a non-starter for Kagi’s ad-free subscription model.[^1]
>Because direct licensing isn’t available to us on compatible terms, we - like many others - use third-party API providers for SERP-style results (SERP meaning search engine results page). These providers serve major enterprises (according to their websites) including Nvidia, Adobe, Samsung, Stanford, DeepMind, Uber, and the United Nations.
The customer list matches what is listed on SerpAPI's page (interestingly, DeepMind is on Kagi's list while they're a Google company...). I suppose Kagi needs to pen this because if SerpAPI shuts down they may lose access to Google, but they may already have utilize multiple providers. In the past, Kagi employees have said that they have access to Google API, but it seems that it was not the case?
As a customer, the major implication of this is that even if Kagi's privacy policy says they try to not log your queries, it is sent to Google and still subject to Google's consumer privacy policy. Even if it is anonymized, your queries can still end up contributing to Google Trends.
I wish there's a RuneScape server for bots. A game where it is both fun to play manually, and automated. Quests, for example, is a one-off thing where it could be fun for the player and not really useful to automate since it is a one-off task (unless you're running bot farm).
I don't think the main game would encourage that - the more obscure the protocol, the less bot in the actual game (even though I don't think it's hard to find protocol documentation or just plug into an official client). OpenRSC, a revival of RuneScape Classic do have botting worlds, but personally RSC is not "fun" for me.
There are programmer games like Screeps (which the new Arena version just launched at the end of last year), but those game usually do not allow manual play or only indirect play. I tried Screeps, but I'm not good at strategy games, so once I get the runtime working I lose interest and none of my friend would want to help me strategize in game that they do not understand.
I don't think this is related to third-party use of CC. This is the same blocking as Anthropic blocking OpenAI access to CC https://www.wired.com/story/anthropic-revokes-openais-access... , which I think is fair that they don't want to help competitors.
Copilot in VSCode is integrated with VSCode's LLM provider API. Which means that any plugins that needs LLM capabilities can submit request to Copilot. Roo Code support that as an option. And of course, there's a plugin that start an OpenAI/Anthropic-compatible web server inside VSCode that just calls to the LLM provider API. It seems that if you use unlimited models (like GPT-4.1) you probably get unlimited API calls. However, those model doesn't seems to be very agentic when used in Claude Code.
Thai banks are required by regulation to have facial recognition when transferring over 50k THB in one transaction or cumulative in a day. I believe most banks have shutdown their internet banking as it's not worth it for the low number of users to implement web-based secure facial recognition that don't allow you to feed spoofed video input. One of the bank that I use will send a push notification to their mobile app for you to confirm the transaction.
I believe that previously internet banking, even before mobile banking, will limit the number of transfer recipients you can add per day/month. With the rise of QR payment I could see this limit being regularly hit if you scrape the web-based banking.
Since the Bank of Thailand claims that they technically don't block many things (mobile banking technical requirements seems to also require blocking root, but they never banned internet banking), I wish there's a new bank that try to disrupt the existing players. But the latest "branchless" banking license were only acquired by existing banking groups, so API-first personal banking remain impossible.