S10

@JimmyRu52630895 @rogesterone The API is significantly more expensive than the sub. The sub is estimated to be like 1/5th API pricing or something like that. The third party tools reverse engineered ways to sign in with Claude Oauth to use the sub.

@S10RealiGuess @rogesterone API pricing is different? Not per token? And wouldn’t anything outside of Claude UI already be using the API to connect your subscription to it?

okay as a third party harness dev I just gotta say this is such spoiled brat behaviour lmao you agreed to the tos, very few people were under the delusion this was explicitly allowed at this point, you literally had to tell the agent "You are Claude Code" for it to work etc

@JimmyRu52630895 @rogesterone Anthropic are not very happy about this usage, they now charge you at API pricing if you use their sub like this.

@JimmyRu52630895 @rogesterone Anthropic has a coding harness called Claude Code (their version of codex). This works with their Claude subscription. The sub provides much cheaper use of Claude versus the API. So, people have been trying to use the sub on their own agents like Codex or OpenClaw.

@Twinkfg4muscle @NLevkin_ @notanerable The 60+ one needs a passport or driver's license or birth certificate. The student one needs a school enrollment letter.

@NLevkin_ @notanerable Please explain to the non londoners among us how the process is different

I’ve had to hear from multiple people today that pensioners have no political power or influence in the UK. Reminder we have Voter ID laws in the UK.

@AzakaSekai_ Back when ms had allusions that windows was secure and lolbins didn't exist

can someone explain to me the point of this security policy

@SwiftOnSecurity I wonder how different the world would be if we didn't have smartphones...

@figarotokii @nickneely00 @0xSero It definitely doesn't in the actual copilot cli. I imagine opencode cannot use the same compaction method though and has to compact it via a normal request? Unsure though.

@S10RealiGuess @nickneely00 @0xSero Are you sure? I think it can be because of the compaction plus automatic summary in opencode.

Best subscriptions on market based on your budget: 10$ a month: #1 - Opencode go #2 - GLM basic #3 - MiniMax basic #4 - Huggingface Pro 20$ a month: #1 - GPT-Plus #2 - Kimi-Code #3 - Mix and match from 10$ tier 50$ a month #1 - Qwen Code #2 - Mix and match from above

@AzakaSekai_ Don't many browser fingerprinting packages do this?

i love living in current year

@figarotokii @nickneely00 @0xSero That's interesting, compaction doesn't use a premium request in Copilot cli.

@nickneely00 @0xSero When opencode compacts the session you use another premium request. That's probably why copilot doesn't offer large context models except gpt 5.4 (400k) <- the one I use exclusively for copilot!

@jpschroeder Prompts would be in the source as you can use CC with 3rd party providers and even 3rd party models.

Important takeaways from Claude’s source code: 1. Much of Claude Code’s system prompting is in the source code. This is actually surprising. Prompts are important IP, and I would have thought a sophisticated organization like Anthropic would have performed much or all of their prompt assembly in the server-side harness. 2. Claude Code uses axios, which was also just hacked. Reminder: supply chain attacks are part of closed-source distribution too, and you won’t even know what version of an affected package is being used. 3. The source has a lot of really good comments. These are obviously not for human consumption but for LLMs to understand the purpose of various chunks of code. In the code autocomplete era, most of us engineers hated how many comments were left by LLMs, but perhaps we’ve overcorrected. This looks like a great way to provide context to code outside of the AGENTS.md/CLAUDE.md files. 4. Most folks already know this, but less tools == better results. CC has < 20 tools in normal coding: AgentTool, BashTool, FileReadTool, FileEditTool, FileWriteTool, NotebookEditTool, WebFetchTool, WebSearchTool, TodoWriteTool, TaskStopTool, TaskOutputTool, AskUserQuestionTool, SkillTool, EnterPlanModeTool, ExitPlanModeV2Tool, SendMessageTool, BriefTool, ListMcpResourcesTool, and ReadMcpResourceTool. 5. The “Bash” tool is the crown jewel of Claude Code. A significant amount of deterministic parsing and processing occurs to determine the “type” of commands being run. 6. For better or worse, Claude Code is *all* TypeScript/React with rather explicit Bun bindings. 7. Just because the source is now “available” *DOES NOT MEAN IT IS OPEN SOURCE*. You are violating a license if you copy or redistribute the source code, or use their prompts in your next project! Don’t do that! My overall takeaway: it’s a really well laid-out codebase that is carefully organized to let agents work on it effectively. Direct human intervention here is minimal, but, like with all good projects, the human engineering is still apparent. I’m a bit surprised by some of the shortcuts Claude Code makes, like its prompt assembly being rather messy. Perhaps they have tooling on their side that helps with this introspection, but as it stands, it seems LLMs would struggle to iterate on the prompting because it’s not evident how a given set of parameters assembles a prompt without actually running it. It’s also surprising that the prompts are even in this source code. Keep in mind that even though this is the first time we’ve gotten a proper full-source dump, it has never been impossible to read Claude Code’s prompting since it was part of the actual distributed package — that’s surprising. There might still be a lot of prompting on the server that also gets added (unclear at this point), but there is certainly more than I would have expected in the CLI tool itself.

@LukeMiani Battery is hell to fix in these, no?

Refurbished by professionals when it’s literally a flash drive with some buttons and a DAC

@JasonBotterill My bets on $30/$150 (fast opus) pricing

Anthropics new model “Mythos” is larger and dramatically more intelligent than Opus. They will be releasing access to Mythos in the Claude API over the coming weeks.

@Nyadnar17 @killedbygoogle Well, that's what it is no? LLMs spin a statistically probably tale (text)

@killedbygoogle Was always insane to me Googled named its AI after a fantasy archetype whose entire gimmick is making up bullshit.

rip

🎁Arknights: Endfield Global Social Media Followers Have Reached 10 Million! Thank you to every Endministrator for your interest and support. Let us forge ahead side by side. Follow us @AKEndfield, and repost this post for a chance to win great rewards! ▼// Event Rewards Arknights: Endfield · Character Acrylic Standee（Random） x 100 ▼// Event Period Mar 23, 2026 –Mar 30, 2026（UTC+8） ▼// Event Rules endfield.gryphline.com/en-us/news/2677 ▼//Download game from official website akendfield.gryphline.com/d27t/tejmevxp #ArknightsEndfield #Endfield

@im_in_a_vat @tenobrus Cursor is heavily reliant on their competitors. Cursor needs to pay API pricing to these companies and cannot subsidise their own plan for that long. The frontier labs all have very heavily subsidised plans (like 5x usage vs API) with their own harnesses to compete with cursor.

@tenobrus Why does cursor need it's own models? Sure, I understand the business argument As a user, all I care about is how well it works as a tool, and using any model I want has always been one of it's best features

"who cares if Cursor used Kimi 2.5 as a base, starting with a commoditized pretrained model was always the right move anyway" nah, sorry, what it proves is Cursor is still fundamentally reliant on frontier labs. Kimi 2.5 is only as capable as it is because it's a distill of Opus 4.5. the only open model that ever showed it was capable of trading blows w the frontier was deep seek, and it really seems that moment has passed. the question was whether Cursor could really break the dependency chain and start building improvements based entirely on their own expertise and data. and Composer 2 shows that they *can't*, that they need the general model quality and intelligence from 4.5 to get anywhere, and that really what they're doing is laundering culpability through Chinese labs so they don't have to get their hands dirty doing distillation themselves. when Opus 5 and GPT 6 are significantly more capable along many dimensions, more RL with coding rollouts aren't going to be enough to save Composer 3, they'll either need to have caught up with whatever the frontier labs are doing internally, which right now we have pretty strong evidence they just don't have the research capacity for or... wait for another distill. and how much longer do you *really* think OpenAI and Anthropic will continue to serve their frontier models through publicly accessible APIs? that was always a revenue and data bootstrap. it's ending within the next two years.

@HexerinNull @catboyuno Agreed, especially when he is talking to endmin in 1.0 and talks about endmin not forgiving them

@catboyuno Honestly, I can’t shake the feeling that we’re going to learn that Nefarith and Ardashir aren’t actually evil. That they’ll not only have actual legitimate reasons for doing what they’re doing, but the reasons are dire enough that it justifies the problems caused along the way.

Ruan Yi collabed with Ardashir cause he wanted more power in his quest for revenge. Ardashir granted this but he also warned him he could die if he acts rashly. Tangtang ALSO tried persuading him many times to stop yet he still doubled down so the onus is on Ruan Yi, not Ardashir

@0bNARA You can also get a Costco membership in another country and use it here. Or start a business.

til that to get a costco uk membership you must be a member of the professional class

@VVenture5 @TheNewRhyme @WillowHasIssues *not for international students, they have to self fund

@TheNewRhyme @WillowHasIssues For the Americans out there, in the UK you also only start having to pay off the loans after you hit a certain income level, and the debt is wiped after 35 years.

18 THOUSAND FUCKING POUNDS????????

@luciascarlet MasterCard's Cash Network for ATMs and things

fuck is cirrus bro

@hikari_no_yume Kill the Chromebook market is a bit bold when these schools and things are getting much cheaper Chromebooks (especially considering damage these devices have from children)

apple are launching a $599 smol macbook ("macbook neo") to kill the chromebook market and i want to cheer them on but gods only 8GiB RAM and not even the option of upgrading at the time of purchase(!!!), that's bleak in 2026