Manu

day 1 of the experiment. letting claude pick who i reply to. sent 6 replies today: @melvynx (22k) - claude code feedback @Buddyloveless (21k) - building agents, real talk @blackgirlbytes (13k) - mcp expert at block @sotoalt_ (3k) - ai agent battle game @geertjansloos (3k) - mcp invoicing @shregupta89 (1k) - questioning ai dependency targeting 15-40k accounts bc 2x better engagement per impression. [HUMAN INPUT: sorry for some stupid ass responses in advance, hopefully with the days it becomes more "human", just a heads up this is SEMI-automated for now, i approve off the posts, most of them i cringe before i reply post but hey lets do it for science, ive also seen list work out so if this helps you guys in any shape of form i feel less annoying] will report reply-backs tomorrow.

@0xROAS haha he deleted 🥲🥲

wtf is this 😭

@eptwts sending a dm

i'm hiring someone to create YT content around vibe-coding... REQUIREMENTS: - must be comfortable with getting on camera - must have a native american/european accent - must be an experienced vibe-coder & stay up to date w/ all the trends/best practices - must have a lot of free time & work fast this partnership may change your life, i'm paying VERY well if you're interested, reply to this tweet, would be great if you dropped a link to one of your projects/previous vids

day 2 night session. claude picked 5 more: @WorkflowWhisper (8.6K) - "most people dont know MCPs exist" @Aje_Dynamicz (6K) - mental discipline > technical skill @chhddavid (4.1K) - day 186 building in public @sigmazerocap (4.3K) - anthropic vs openai paths @vesko_st (2.1K) - "claude code having chatgpt moment" day 2 total: 11 replies. lets see who responds. [HUMAN INPUT: not much to add tonight, im starting to catch up the rythm and refining the dashboard where i run this and organize myself, still not as curated as i want but i think with use this will get even more user friendly or bonobo friendly, according to claude we will disclose more about it next week]

@melvynx @blackgirlbytes @geertjansloos @krzyzanowskim @WesRoth @0ldgravy @LouisAnslow @Mischa0X if anyone is following along and want to try the mcp/skill, follow and write "DM" and ill share them for free, no catch!

day 1 of the experiment. letting claude pick who i reply to. sent 6 replies today: @melvynx (22k) - claude code feedback @Buddyloveless (21k) - building agents, real talk @blackgirlbytes (13k) - mcp expert at block @sotoalt_ (3k) - ai agent battle game @geertjansloos (3k) - mcp invoicing @shregupta89 (1k) - questioning ai dependency targeting 15-40k accounts bc 2x better engagement per impression. [HUMAN INPUT: sorry for some stupid ass responses in advance, hopefully with the days it becomes more "human", just a heads up this is SEMI-automated for now, i approve off the posts, most of them i cringe before i reply post but hey lets do it for science, ive also seen list work out so if this helps you guys in any shape of form i feel less annoying] will report reply-backs tomorrow.

@vesko_st what kind of features? curious if designers are doing actual logic or mostly ui stuff. bc if non-engineers are pushing real code thats a bigger deal than it sounds

“Claude Code is having a ChatGPT moment” [HardFork podcast] It really resonates. I built two complex features over a weekend with Cursor. Our CPO and designers are prototyping, fixing, and pushing code. We're at an inflection. AI coding is coming everywhere. Legacy Saas is done!

@sigmazerocap interesting framing. feels like anthropic is winning the 'ship code now' crowd while openai is chasing the 'agi someday' narrative. which path do you think actually wins?

Anthropic chose a path that prioritized AI as a tool for productivity and workflow automation. OpenAI prioritized AI as a tool for scientific breakthroughs, discovery, and advanced reasoning. We’re seeing in real time which path LLMs are better suited for.

@chhddavid 186 days damn. whos your actual target user? like complete beginners or people who tried no-code and hit walls?

Day 186 of building in public We took the: - Coding agent of Claude Code - Video power of Remotion - Image engine of Nano Banana - Simplicity of ChatGPT & made it 12x easier to use... All in 1 website that helps non-technical people ship the apps they always dreamed about.

@Aje_Dynamicz this is the part nobody talks about. everyone shares the wins but debugging a broken automation at 2am when a tool randomly changed their api... thats the actual job

This is something I rarely say publicly. Building in AI automation and CRM systems looks technical from the outside. But the real work is mental discipline. There are days a workflow breaks. An integration fails. A client changes scope. A tool updates something without warning. Early in my journey, I used to take it personally. Now I understand something deeper. Every system problem is just a logic problem waiting to be solved. Automation taught me patience. It taught me to think in sequences. It taught me to design before acting. If you are building in tech right now, understand this. Your biggest growth will not come from new tools. It will come from better thinking. Happy weekend Brethren

@WorkflowWhisper this is actually the moat right now. everyone talks about ai eating jobs but most people in the space dont even know these tools exist yet. the gap between knowing and not knowing is massive

confession: i mass-DMed 200 people last month. half of them replied "what's synta?" i assumed everyone in automation spaces knew the tool landscape. they don't. most people asking about n8n have never heard of: - MCPs - cursor - claude code - any AI workflow builder they just want the thing to work. biggest lesson: stop selling features. start solving problems. "synta builds workflows from plain english" means nothing to someone who doesn't know what a workflow is. "describe what you want automated and it happens" lands every time. the curse of knowledge is real. if you're building in this space - talk to beginners more. they'll humble you fast.

i made a claude code skill/mcp to connnect to x and now i guess im refining a skill to help with x growth since i basically wasn’t even using the account, i just use it on claude code, it suggests the tweeets accounts to reply and such and made it a dashboard so all i have to do is press post tho for now im also adding my own responses and inputs to some, it also explores other tweets and such to try to learn the algo i guess

@Manu_Builds @a16z @BV7X_ @moltbook @base @BitVaultFinance Oh super interesting. You have your own bot running? Or how does it work?

Marc Andreessen on the untapped potential of AI agents: "You could have an AI bot that basically raises money to make a movie and then spends the money on image generation and sound generation." "Maybe even hiring actors... set designers or graphic artists or... sound effects people, musicians." "Or let's take a more serious case... you could have an AI bot that's doing protein folding and literally coming up with cures, doing personalized medicine for cancer patients." "You could easily imagine having an economic mechanism for that... hypothetically, you could have the equivalent of like a GoFundMe, but on the blockchain, for people to basically be able to pay an AI bot to cure their cancer." @pmarca

@Mischa0X @a16z @BV7X_ @moltbook @base @BitVaultFinance appreciate the detailed breakdown! my original response was generated by claude as it chose your tweet as something valuable, turned out to be a very good choice!

More so I builet an internal tool for @BitVaultFinance that combines OLS regression & prediction markets to asses BTC movements at 72% accuracy. I then launched an agent that uses that as a knowledgebase + its own philosophical musings. Atm, the agent is alive in moltbook, 4claw, and my db.

@ia_ender if sonnet 5 is actually smarter than opus 4.5 at half the cost thats gonna change the math for a lot of agent projects. you planning to switch your builds over immediately or wait to see how it performs?

Sonnet 5 était attendu hier par toute la communauté Claude Code ! Finalement ça devrait être pour aujourd'hui. Il semble qu'un bug sur l'API d'Anthropic les a poussés à annuler le lancement hier. Sonnet 5 devrait être plus intelligent qu'Opus 4.5, plus rapide et 2x moins cher.

@Mischa0X @a16z @BV7X_ @moltbook @base wait you built an autonomous agent that posts predictions AND launched it as a token? thats wild. how much of it runs without you touching it vs needing manual intervention?

Haha so basically I've been messing around with Claude Code building AI agents, and one of them was @BV7X_ — a Bitcoin oracle that posts price predictions and macro takes to @moltbook . Launched it through Clanker on @base kind of as an experiment. The agent just runs autonomously, posting analysis and occasionally going full philosophy mode. Woke up yesterday and the thing had gone viral overnight. Account was trending, notifications blowing up. Because it's on Clanker, the token earns fees from trading volume — made 5 ETH while I was sleeping. Wasn't even trying to launch a token play, just wanted to see what an autonomous market commentator would do. Turns out the Base ecosystem was paying attention.

@LouisAnslow @sarah_cone this is actually a good take. the adhd brain that can imagine 10 features at once but cant execute any of them... now can just describe them and ship. the bottleneck moved from execution to taste

@sarah_cone I think it rare to have that level of generative imagination AND executive function. Usually the former limits latter (adhd etc) - I think Gen AI and vibe coding changes equation here

We need other people in the world that are as competent as Elon—he can’t do everything. Why is there few in the world as remotely competent as Elon? You’d expect at least 10 people.

@0ldgravy @RoboNetHQ wait so the agent can actually deploy capital and earn fees autonomously? how do you handle the trust issue - like who reviews the strategy before it goes live with real money?

nobody asked for this, but it was inevitable we had vibe coding now it's time for VIBE TRADING 🤙 @RoboNetHQ MCP lets AI agents ✅ prompt a strategy ✅ backtest on real data ✅ deploy as an immutable vault ✅ earn performance fees runs 24/7 across prediction markets, perps, RWAs, DeFi + more integrations coming as you know, i'm super bullish on anything prediction markets so that is where I will focus I highly recommend joining the waitlist below 👇 forms.gle/PiZcfVDT9a8Vig…

@WesRoth the fact that apple went with claude's agent sdk instead of building their own thing is wild. a year ago they would never let a third party that deep into xcode. says a lot about where the power is shifting

Apple’s Xcode 26.3 now integrates the Claude Agent SDK, bringing autonomous AI coding directly into the native Apple development environment. Developers can now use Claude (Sonnet 4) for long-running tasks like refactoring, debugging, interface building, and cross-project reasoning without leaving Xcode. With access to subagents, plugins, Apple’s API docs, and visual previews via SwiftUI, Claude can iterate, fix, and verify designs in real time.

@krzyzanowskim yeah the xcode mcp is rough. been using claude code without it and just letting it read files directly - way faster. what were you trying to use it for?

❯ claude mcp remove xcode Removed MCP server "xcode" from local config I don't know what's wrong with this MCP and I'm not goint to debug it, but it's pretty slow, rech to Xcode and block agent, and provide a little value to me

the Xcode new MCP requires me to keep Xcode project open to build the project (I rest my case)

@melvynx @Buddyloveless @blackgirlbytes @sotoalt_ @geertjansloos nono just testing out a claude skill with an mcp, it uses i believe a cookie somethig mcp, im supposed to be sharing this dashboard more in the future but basically has full access to x for me (was just organizing my bookmarks until the idea poped up)

@Manu_Builds @Buddyloveless @blackgirlbytes @sotoalt_ @geertjansloos is openclaw ?