Ashley Rodan

@build_aus Fairdinkim someone ought to do this 💚💛

Day 1 and we've rocketed to 1,000 followers. That's 1,000 people that support a movement and mindset shift to shake this country out of stagnation. Thank you all. But we're just getting started. #buildaustralia

@mattpocockuk Minimum two agents per coding session. Mainline (right pane) runs the inner loop. Plans, builds, updates its spec with relaxed perms. I grill the diffs and test the output. Outer loop (bottom pane) optimizes the mainline. Tunes the harness, refines context and harvests skills⚒️

What 'advanced' AI coding techniques are you using? I.e. what do you feel like you've discovered that no-one else knows about yet?

@itskerrow @theo Need more info here this sounds great

telegram. supergroup gets you threads, one session per thread. pin a few threads to specific agents. add control functions to bypass agent and execute specific control actions. run requests async (respond immediately with "..." then edit message every 2s with tool calls / response). transcribe voice messages. allow images in/out. take advantage of reactions and replies. bonus: allow parallel concurrent requests by forking session and reconcile returns by sending response into active session (most recent returned fork wins and replaces active).

I want to control my agents from my phone. I'm sure you do too. What workflows do you want? Do you have one that works now? Tell me all about it.

@benslockedin unreal how'd i miss this. Do you have it hosted?

last year i rebuilt a $96 million government website in an hour - it took deloitte seven years - that's the whole tweet man, I still just find this really funny

@mattpocockuk Is a skill going to block it do we need a hook?

I get a lot of questions about how I stop Claude Code running dangerous git commands. The secret is hooks. So, I packaged it up into a skill. INDISPENSABLE when running Ralph in a docker sandbox. Get it here: aihero.dev/s/jiOinX

@akshay_pachaar Playbook for hosting it. Clawdbot on railway has not been working too well thinking of the cloudflare sandboxes

The best AI projects are the ones you can actually understand. nanobot is an open-source personal AI assistant built in just ~4,000 lines of code. That's 99% smaller than the project it's inspired by (Clawdbot has 430k+ lines). Most AI agent frameworks are bloated and impossible to navigate. nanobot takes the opposite approach and delivers core agent functionality with a codebase you can read on a weekend. Key features: - Works with any LLM provider - Run local models with vLLM - Connects to Telegram, Discord, and WhatsApp - Ships with scheduled tasks, persistent memory, and a skills system Everything you need for your personal AI assistant. You can get it running in 2 minutes: 𝗽𝗶𝗽 𝗶𝗻𝘀𝘁𝗮𝗹𝗹 𝗻𝗮𝗻𝗼𝗯𝗼𝘁-𝗮𝗶 𝗻𝗮𝗻𝗼𝗯𝗼𝘁 𝗼𝗻𝗯𝗼𝗮𝗿𝗱 𝗻𝗮𝗻𝗼𝗯𝗼𝘁 𝗮𝗴𝗲𝗻𝘁 -𝗺 "𝗛𝗲𝗹𝗹𝗼!" Here's something worth thinking about: The LLM alone is not what makes an agent powerful. It's the harness around it: tool execution, memory, context building, and task scheduling. That's where the real engineering lives. Learning to build effective agent harnesses is quickly becoming one of the most valuable skills you can develop. nanobot gives you a clean, minimal codebase to study exactly that. That said if you want to understand the full picture before diving into this lightweight version, I also recorded 30 minute video on mastering OpenClaw/Clawdbot, the project nanobot is built on top of. I have linked the tutorial in the next tweet.

@r_vkrm @openclaw @AnthropicAI This is a perfect troll you might have copyrights to this when it happens

@openclaw @AnthropicAI in two weeks time😭

🦞 BIG NEWS: We've molted! Clawdbot → Moltbot Clawd → Molty Same lobster soul, new shell. Anthropic asked us to change our name (trademark stuff), and honestly? "Molt" fits perfectly - it's what lobsters do to grow. New handle: @openclaw Same mission: AI that actually does things.

@KingBootoshi What interface is this. Have you got details on your setup?

having claude code create their own CLI tools then save it to his local knowledge bank to refer in future sessions IS STUPID OVERPOWERED ‼️ my cc is evolving through usage and ADAPTING to my own workflow overtime those who don't use now, are missing out on crucial growth!

@benslockedin Who writes human.md is humans for humans or ai for humans?

@gioragutt @grok @NikhilAdit25502 @supabase So @ElectricSQL ?

@grok @NikhilAdit25502 @supabase @grok so… cdc?

We've released something new to the Postgres world: asynchronous streaming between databases 🎄

@serafimcloud Claude android has code support tell me how this is different

What if your Claude Code client worked beautifully on mobile? Same calm UI. Multiple branches. Live previews. And the ability to keep working without your laptop. Free to use with your own Claude subscription Reply if you want early access

@abhitwt TDD

i am a Vibe Coder, scare me with one word

@GeoffreyHuntley Yeah man you've triggered a revolution

Hahahaha holy s… There’s an official plugin in Claude code for Ralph now

@mattpocockuk How do you rate copilotiit and the agui standards

AI SDK v5 Crash Course dropped yesterday One detail I missed off the announcement: the first 2 sections are FREE. That's 15 lessons of goodness you can be enjoying right now. Check 'em out: aihero.dev/workshops/ai-s…

@JahnaviBhansal1 Not many people chatting about this! Mind slightly blown! Screenshot of me fishing through the trash 😂

“Imagine with Claude” feels like a Generative OS that builds software on the fly with every click. Told it to make a desktop of a founder who promised to ship by Friday

@GeoffreyHuntley Any insight to your setup as to how you had it running for 3 months??

Re: costs Technically it costs about 5k usd to build your own compiler now because cursed was implemented first in c, then rust, now zig. So yeah, it’s not one compiler it’s three editions of it. For a total of $14k USD.

Verifying myself: I am ashrodsan on Keybase.io. dSE74PlFvr28nhXOMxMNsWY2gAdHJ7GoyBSS / keybase.io/ashrodsan/sigs…

@nickbaumann_ This area is interesting do you have more details. It sounds You're just comparing cursor to Cline. What about Claude code?

Cline optimizes for maximum capability: direct model access, full context, zero throttling. This often costs more, but the productivity gains can be exponential.

No, @cline is not 'open-source Cursor' or 'open-source Windsurf.' It's a fundamentally different approach to AI coding. more thoughts below

@lbeurerkellner @marco_milanta Can you confirm how the agent is managed. it's clearly not using the official GitHub integration that Claude code ships with.

😈 BEWARE: Claude 4 + GitHub MCP will leak your private GitHub repositories, no questions asked. We discovered a new attack on agents using GitHub’s official MCP server, which can be exploited by attackers to access your private repositories. creds to @marco_milanta (1/n) 👇

Blown my #claudecode usage in a 1.5 hour sitting with ~14M input tokens used. #Claude4 is impressive but claude code prompts wastage and unexpected rate limits makes it difficult to rely on.