𝗥𝗬𝗔𝗡 𝗟𝗘𝗘

@alkimiadev How did u even get access. they are consistently sold out daily

Yeah no complaints so far. I don't work on stuff that would trigger censorship or would offend the ccp but it is probably worth keeping in mind that these models are more heavily censored on subjects considered sensitive by the Chinese government. If you're just working on every day dev stuff that shouldn't be an issue but if you're doing any nsfw, certain types of crypto work, and a few others then it almost certainly would be an issue.

After getting frustrated with zai serving a bricked quantized model for the past week I decided to try out the new Alibaba's new pro coding sub. It comes with several open models including GLM 5. I might cancel the zai sub if they don't fix it asap

@alkimiadev @Zai_org @grok Yeah absolute joke ATM. Compare and constrast GLM-5/GLM-4.7 On Alibaba coding plan versus z.ai coding plan

I'm considering canceling my @Zai_org subscription over them bricking GLM 5 AND raising prices. I just can't justify paying a quartery subscription when the main model reliably flips into "gibberish mode". They jokingly(had to be a joke right?) called it "degraded model performance" but it is full on gibberish an totally unusable. I've reverted to using GLM 4.7 because of this. I'm fine with working with GLM 4.7 since I was every day before they released GLM 5. GLM 5 is a much better dev, even the heavily quantized version before it shifts into gibberish mode, but it is entirely unreliable right now. The second screenshot shows an example where they slipped into gibberish mode in the middle of a task. They corrupted a file which caused me to have to revert their changes. I ultimately had to switch back to GLM 4.7 to complete these tasks. I'm thinking of switching to Alibaba's coding plan since it offers several open models and includes GLM 5(hopefully not a broken quantized version like at zai). Regardless there is absolutely no way I'm going to pay the increased fees for the quarterly max subscription if the main model, and the reason for the price increase, is totally bricked like this.

super thankful I got to meet one of our top users ever: Ryan has been using Magic Patterns for 2.5+ years!! He is the reason we have core features like "forking" (his feature request 2 years ago) and he created a Magic Patterns MCP before it was a thing!!!

@daniel_mac8 @garrytan dont forget SQL

Manus ex-backend lead had a genius insight text based clis beat structured tool calling for ai agents all day because unix commands appear in training data going back to the 1970s text is the native language of the command line AND text is the native language of llms

just fyi if you're hitting segfaults with claude code, you can run it with deno and also get the benefit of less memory usage

@maiahmac @astrodotbuild Explain how this works for us mortals?

Simple Astro component that extends the built-in ClientRouter to support client-side page cache. Then combo this with Astro's Server Islands and you get the same behavior w/ next.js Cache Components 🚀 @astrodotbuild #astro #astrojs astro.js

@sebiomo_ Creating a security skill and then recommending people to run it with npx. The irony

I shared my security.md template last week. 80K+ people saw it. 1,000 bookmarked it. So I turned it into a tool. npx secure-repo audit - scans your repo for security issues in 10 seconds. npx secure-repo init - drops production-grade security policies into any project. Free. Zero dependencies. One command. The problem: AI agents (Cursor, Claude Code, Copilot) write code that works - but without guardrails, it's not secure. No input validation. Secrets in client code. No rate limiting. A SECURITY.md in your repo root changes that. The agent reads it and follows the rules. The free version gives you 3 files: SECURITY.md - secrets, access control, incident response AUTH.md - tokens, sessions, passwords, roles API.md - validation, rate limiting, error handling Every rule is marked "MUST FOLLOW." Every pattern is copy-pastable. Don't use the terminal? Just tell your AI agent: "Run npx secure-repo audit in my project" It runs the scan, reads the output, and fixes the issues. Works with any agent. Want complete coverage? The Pro Pack adds 27 more files: Database, deployment, incident response, payments, data privacy, rate limiting, access control, a 100+ point audit checklist, Supabase & Firebase presets, and code examples. $19 launch price (3 days only). Then $29. shipsecure.sebiomo.com

@kvnwolf @mattpocockuk Use oxlint with type aware and typecheck on

@mattpocockuk I recently moved to tsgo as well and added a post tool use hook to run tsgo after every file edit (helps Claude catch errors early). Combined with turbo (with remote caching enabled), incremental compilation and tsbuildinfo makes it really fast.

I'm in the phase of my Ralph journey where I'm happy with the quality and now optimising for speed tsc -> tsgo Task caching with turborepo Let's get SHIPPING

@Marsmensch GLM 5 cannot see like GLM 4.6v can. Are you actually getting good results?

FRONTEND UI/UX => zai-coding-plan/glm-5 Builds gorgeous UIs. Ideally this would be Gemini 3 Pro, but I am not convinced of the models in general. GLM 5 is next in the official visual-engineering fallback chain (Gemini → GLM-5 → Opus → Kimi). Solid substitute. I can always use a dedicated tool for UI/UX if need be.

I spent way too long optimizing my @opencode + oh-my-opencode model config in the past days to find out IF I could do without Anthropic. Turns out I can! Codex, GLM & MiniMax are perfectly balanced now for my use cases and I will cut down one more subscription next. Full breakdown to save you some time

@justsisyphus @sperosck @AnthropicAI What is the full recommendation if no anthropic models? KIMI 2.5 -> Sis GLM 5 -> ULW Sis Junior -> ????

@ryanleecode @sperosck @AnthropicAI sisyphus works fine on glm too tho

yes we rewrote our ultralong readme and we fixed our readme: we don't say much of @AnthropicAI models, too we don't want people's account to be banned now we have decent gpt support, opensource model support, hashline edit from oh-my-pi support by default hope you can ultrawork better

@justsisyphus @sperosck @AnthropicAI No GLM?

@sperosck @AnthropicAI sisyphus -> kimi k2.5 hephaestus -> gpt 5.3 codex medium prometheus, atlas -> if it is not gemini or qwen it is okay

What is even the point of the When to use this Skill section. it serves no purpose as this AI has already ingested the skill

My next guest, @EnoReyes (co-founder of @FactoryAI), built this AI PM skill that blew my mind. It includes: → Product principles and positioning → What an 11-star experience looks like → PRD templates and review rubrics → Prioritization frameworks ...and more It's 700+ words and I'll admit it made me a little worried about my profession 😅 Subscribe to get our interview and Eno's full skill tmr: @peteryangyt?sub_confirmation=1" target="_blank" rel="nofollow noopener">youtube.com/@peteryangyt?s…

@SimplySigmoid @tom_doerr im afraid this needs to be deleted immediately

@tom_doerr I prefer to have 100% control over my data and ingest it into neo4j all entities and relationships between then are also 100% manually managed & improved

Framework for building knowledge graphs from text github.com/Hawksight-AI/s…

@EffectTS_ @DrizzleORM @claudeai @polar_sh @kitlangton Juicy

This week in Effect: → @DrizzleORM now has native Effect Schema support! 🎉 → @claudeai keeps getting Effect-pilled 🤖 → @polar_sh using Effect in production! ✨ → Effect and the Near Inexpressible Majesty of Layers by @kitlangton More below: effect.website/blog/this-week…

🧪 Experimental: Use OpenCode with Claude Code, Codex, and Amp - Universal coding agent control - HTTP API for sandboxed agents - OpenCode TUI, web UI, SDK Available in Sandbox Agent SDK 0.1.6

Our internal lint rules are now open source, featuring 23+ custom rules we use to guide droids at @FactoryAI. These rules cover file organization, React patterns, testing, error handling, API conventions, and more. Every rule is 100% droid-generated and includes detailed Markdown docs. While this isn't meant to be imported as-is, we hope it inspires you to build custom linting rules tailored to your own codebase. Take what works, ignore what doesn’t, and tweak as needed. You can leverage the docs as building blocks to suit your specific framework or language. Check it out: github.com/Factory-AI/esl…

@Shelpid_WI3M delete this account

@justsisyphus @techscienceandy @opencode @OpenAI do i need to disable the default plan and build agents?

@techscienceandy @opencode @OpenAI sisyphus + ulw = i don't want to think prometheus + atlas = i know what i am doing and i want my agent to work exactly as i have planned hephaestus = i know the what would outcome be like but i don't want to care about the details

Meet another agent on @opencode The wait is over, the Sisyphus moment, but for @OpenAI Hephaestus, on oh-my-opencode Hephaestus succeeded in the task of improving test execution time that Sisyphus failed, reducing it to one-third with just a single prompt. Soon, guys!

@davis7 @vercel good luck if you're using worktrees

The "just-bash" package is the best thing @vercel has shipped in a long time Insanely useful for custom agent stuff