Mo Vaid 💙

@_watzon @tommoor Supports fn hotkey with PTT?

Just gonna drop this here because the algorithm keeps showing me these posts. Pindrop is a 100% mac-native alternative to Handy with better performance and more features. All it's missing at this point is signed builds because I can't afford an Apple Developer account 😅 github.com/watzon/pindrop

If you're paying for dictation software... why? handy.computer is free, includes excellent open source models, runs LOCALLY, works exactly the same as those apps charging a subscription and sending everything you say to the cloud.

@MKumarTweets @hasantoxr @grok "just" Clowns

@hasantoxr @grok when was LangWatch first available and what is the delta of features since then

🚨 BREAKING: Someone just open sourced the missing layer for AI agents and it's genuinely insane. It's called LangWatch. The complete platform for LLM evaluation and AI agent testing trace, evaluate, simulate, and monitor your agents end-to-end before a single user sees them. Here's what you actually get: → End-to-end agent simulations - run full-stack scenarios (tools, state, user simulator, judge) and pinpoint exactly where your agent breaks, decision by decision → Closed eval loop - Trace → Dataset → Evaluate → Optimize prompts → Re-test. Zero glue code, zero tool sprawl → Optimization Studio - iterate on prompts and models with real eval data backing every change → Annotations & queues - let domain experts label edge cases, catch failures your evals miss → GitHub integration - prompt versions live in Git, linked directly to traces Here's the wild part: It's OpenTelemetry-native. Framework-agnostic. Works with LangChain, LangGraph, CrewAI, Vercel AI SDK, Mastra, Google ADK. Model-agnostic too OpenAI, Anthropic, Azure, AWS, Groq, Ollama. Most teams shipping AI agents have zero regression testing. No simulations. No systematic eval loop. They find out their agent broke when a user tweets about it. LangWatch fixes that. One docker compose command to self-host. Full MCP support for Claude Desktop. ISO 27001 certified. 100% Open Source. (Link in the comments)

@mahadegg @yazins github.com/yazinsai/offli… repo is open

@yazins is this going to be a paid app? Any plans to write a post to teach us?

✨ now available for you to try: offline-tarteel.whhite.com try turning off the wifi after loading the model 😎

@AIGuy144664 @GoogleAI Cuphead vibes! What was the prompt?

@GoogleAI 1930 but fluffy and colorfull.

To kick off your creativity, here are some Nano Banana 2 prompts to try out. We hope you find them fruitful 🍌🍌 Prompt: Create a funny 4-part story featuring 3 fluffy creatures building a treehouse. The story has emotional highs and lows and ends in a happy moment. Maintain consistent identity across the 3 characters. Generate 4 images in 16:9 format, one at a time.

@MrAhmadAwais @shadcn x.com/i/status/20271…

@shadcn Not $2K? 😆

We're a few months away from a $500/month plan. maybe weeks.

@shadcn @FactoryAI had an Ultra plan @ $2000/m but that's seemingly been discontinued

@brian_lovin @ImSh4yy x.com/i/status/20267…

@brian_lovin @ImSh4yy No need to, as of today.

I'm getting close to a year running TanStack Start in production. I have two apps on Cloudflare Workers, they are completely stateless and talk to the backend API via tRPC. I'm not using any of the proprietary features of Workers, just have their Vite plugin to serve the Start app. Everything else is containerized and deployed to Swarm. It's been a great experience so far, and I can't recommend it enough. If you're starting a new project, or looking to migrate your current one to a different React framework, I highly recommend giving it a try.

@yishi969866 @SaltlightHill @UnslothAI @Alibaba_Qwen How would I quantize it myself?

@techMosavvy @SaltlightHill @UnslothAI @Alibaba_Qwen Use INT16 without quantization. You can quantize it yourself—the precision might decrease slightly, but it will use significantly less VRAM.

🚀 Introducing the Qwen 3.5 Medium Model Series Qwen3.5-Flash · Qwen3.5-35B-A3B · Qwen3.5-122B-A10B · Qwen3.5-27B ✨ More intelligence, less compute. • Qwen3.5-35B-A3B now surpasses Qwen3-235B-A22B-2507 and Qwen3-VL-235B-A22B — a reminder that better architecture, data quality, and RL can move intelligence forward, not just bigger parameter counts. • Qwen3.5-122B-A10B and 27B continue narrowing the gap between medium-sized and frontier models — especially in more complex agent scenarios. • Qwen3.5-Flash is the hosted production version aligned with 35B-A3B, featuring: – 1M context length by default – Official built-in tools 🔗 Hugging Face: huggingface.co/collections/Qw… 🔗 ModelScope: modelscope.cn/collections/Qw… 🔗 Qwen3.5-Flash API: modelstudio.console.alibabacloud.com/ap-southeast-1… Try in Qwen Chat 👇 Flash: chat.qwen.ai/?models=qwen3.… 27B: chat.qwen.ai/?models=qwen3.… 35B-A3B: chat.qwen.ai/?models=qwen3.… 122B-A10B: chat.qwen.ai/?models=qwen3.… Would love to hear what you build with it.

@codyschneiderxx generator

I just had Claude Code build me a Facebook ad generator that can make 100+ on-brand ad variations in minutes for $0. And I made a full Notion document guide for you. It includes: 1. How to use Claude to find the pain points and desired outcomes of your ICP 2. How to use these pain points and outcomes to write ad copy variations 3. How to build a Facebook ad template entirely with code (just like the ones you see) 4. How focus Claude Code’s design so the ad feels “on-brand” 5. How to export the Facebook ads as PNGs in a zip file 6. How to bulk upload them to a Facebook ad set 7. How to use an AI data analyst to track the success of these ads Everything above is just API calls and Claude Code doing the work for you. You just come up with the ideas and polish the outputs. Like and comment "generator" and I'll send the Notion document to you

@SaltlightHill @UnslothAI @Alibaba_Qwen Thanks, that's actually not as terrible as I thought it would be. Do you happen to know which quantization level that was at?

@techMosavvy @UnslothAI @Alibaba_Qwen Someone was saying 70

@UnslothAI @Alibaba_Qwen at what TPS?

@Alibaba_Qwen Amazing release! You can run Qwen3.5-35B-A3B locally on just a 24GB Mac / RAM device via GGUFs! huggingface.co/unsloth/Qwen3.…

@YuminAI @matholive1ra @hit_and_trial @chang_defi You won't know it's AI

@matholive1ra @hit_and_trial @chang_defi How?

HAHAHAHAHAHAHAHAHAHAHAHAHAAHAHAHAHAHAHAAHAHAHAAHHAAHAHAHHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAZHHAHAHAHAHAHAHAHAHAHAHAAHAHAHA

@_toddanderson @openclaw How would it achieve something like route optimisation?

Lead Capture & Speed to Lead: 1. Instant SMS response to web forms (<2 min reply time) 2. Quiz funnels that qualify leads before human touches them 3. After-hours lead capture (texts prospects "we'll call you at 8 AM") 4. Missed call auto-text ("Saw you called, what do you need?") Scheduling & Dispatch: 5. Client self-service booking agent (eliminates dispatch agent entirely) 6. Dynamic tech matching (AI picks the right tech based on job type, location, past success rate) 7. Clock-in/out monitoring via SMS ("text START when you arrive") 8. Route optimization (pack more jobs per day, less drive time) Customer Service: 9. Payment reminder texts (auto-chase invoices) 10. Review request automation (send at optimal time based on job completion) 11. Rebooking nudges for recurring services (6-month gutter cleaning reminders) 12. ETA updates ("Juan is 15 min away") Marketing & Content: 13. GBP post automation (weekly updates, photos, promos) 14. Social media posting (Facebook + Instagram, branded content) 15. Before/after photo galleries (auto-generated from field photos) 16. Seasonal campaign automation ("Spring gutter cleaning sale") Operations: 17. Job photo verification (techs upload proof of work, AI checks completeness) Note: this one is a work in progress. 18. Upsell prompts (CRM detects patterns: "This client always says yes to add-ons") 19. Supply reorder alerts (predict when you'll run out of parts) 20. Crew performance tracking (which tech closes the most upsells?) Financial: 21. ROAS analysis across all channels (Google/Meta/Yelp/Angi — which actually pays?) 22. Pricing intelligence (scrape competitors, suggest rate adjustments) 23. Cash flow forecasting (based on job pipeline + historical payment speed) Conversational Backend (zero UI needed): 24. Slack/WhatsApp-based CRM (no login, just text "show me today's jobs") 25. Voice dispatch ("Hey Siri, who's available for an emergency call?") 26. Natural language reporting ("How many jobs did Juan close this month?")

Heres a list of 26 things your @openclaw agents can do for your blue collar business that we're building for our clients:

@whoiskatrin @jachands Is it kept up to date with recent SDK changes?

@jachands examples is a good place to point at, for sure

🎈Agents SDK v0.5.0 is out! We’ve fully re-written @cloudflare/ai-chat with data parts, persistent tool approvals that survive page refresh and DO hibernation, and maxPersistedMessages to cap SQLite storage More news things: • built-in retries with exponential backoff • protocol control per connection • simpler queue and schedule APIs • persistent chat state and storage limits • major reliability fixes across streaming and hibernation go make cool stuff with it!

@Braders2k @code_rams That's yearly but yeah within a rolling 5hr window. Also you get 10% off with a referral (mine below) platform.minimax.io/subscribe/codi…

@code_rams Started plan is $100 and says 100prompts/5hrs? -

Been using MiniMax M2.1 as my primary for weeks now! here's why i keep going back: → writing style just clicks with what i need → context stays solid through long messy threads → 300+ prompts in 5 hours on $20/m? easy → no token jail, no surprise bills Handles 90% of my daily work: Chiti routing, planning, light coding, agent stuff when things get properly complex? fallback to Opus and Codex save the build - cheap, fast, reliable, huge context and now they just dropped M2.5 80%+ SWE-Bench Verified 37% faster on complex tasks SOTA agent/tool calling same low-cost vibe if M2.1 already felt this smooth? M2.5 might make Opus/Codex fallbacks rare $20/month this has been the sweet spot in my multi-model setup if you're tired of burning credits? worth giving the MiniMax coding plan a shot Happy to learn from you! who's already testing M2.5? drop your real usage below

@genzvibefounder @pnegahdar > minis Plural? 👀

@pnegahdar Easy to set up with local models/lm studio? I blew my OpenAI budget on Mac minis 💀

We built an AI coding tool as a side project. It was building itself in days. After few weeks, it replaced our IDE, terminal and todo lists. We now spend 90% of our screen time in it. Today we're open-sourcing it. It's called OpenADE and here’s a quick breakdown. So, we’ve been using Cursor, Claude Code and Codex like eveyone. Similar 6-step workflows: 1. agent writes code 2. you squint at a diff 3. you accept it 4. you chug a sugar-free Red Bull 5. something breaks 6. you back and forth 100 times. The tools were fast but the results were sloppy…and took a lot of cleanup to make sure we weren’t shipping the slop. While planning helped, we also found issues working Claude Code . There was no way to comment on a plan. and no way to refine it without starting over. Just endless back and forth in a chat window. So we built the entire OpenADE tool around it: Plan → Comment → Revise → Red Bull → Execute. We made sure the plans aren’t a vague summary. It tells you what files change, what interfaces break, what key decisions were made and why. Your teammate can review it before a single line of code is written. Comment on it like a PR. Catch the bad architecture call at minute 2 (not hour 2). OpenADE is MIT licensed and runs locally: > No cloud > No account > No per-seat pricing > Uses your Anthropic API key or Claude subscription. It's really great. Star it, try it, break it: github.com/bearlyai/opena…

@crazibeat1 @dart_lang @FlutterDev Looks cool. GitHub link?

🚀 Just published: sequelize_orm The most type-rich, autocomplete-friendly ORM for @dart_lang & @FlutterDev ✅ Code-gen gives you compile-time safe queries, models & associations ✅ Supports Postgres, MySQL, SQLite, MariaDB, MSSQL Check the thread for a full tour! 🧵

The best way to learn how something works is to build it yourself. So I rebuilt OpenClaw from the ground up, starting from a 20-line Telegram bot and ending with a Mini Openclaw in 400 lines. I learned a lot and it was a lot of fun! Here's the tutorial so you can do it too:

@levelsio @digitalshane_ consumer.org.hk/en/press-relea…

@digitalshane_ No not everything Steel, glass, wood, stone, ceramic is generally fine

It's not just paper It's thermal paper which is coated with BPA and BPS, endocrine-disrupting chemicals that are absorbed through your skin There's more BPA in a single piece of thermal receipt paper than an entire plastic bottle used for years Repeated exposure to these chemicals are linked to infertility, metabolic dysfunction, and hormonal imbalance