Alejandro Peghin

If @openclaw is your agent, SupraWall is the law. 🛡️Open-source runtime firewall for AI agents. 🔒Vault secrets. 🤑Cap budgets. ⛔️Block unsafe actions. EU AI Act ready. → supra-wall.com

Launching next week our native integration with @Paperclip_AI All your agents fully protected by our open source security #paperclip

SupraWall's blueprint shuts AI agent loopholes. Probabilistic security is a failure. #AIAgents #CISO

@tamir_eden Claude handling brand ops like canceling orders and running fraud queues is early but real

This one is for the brand operators! We gave Claude all the skills you need for running opeartions with ease. ▎ • cancel & restock orders ▎ • bulk-update 10k product prices ▎ • run a fraud review queue ▎ • calculate repeat purchase rate by product ▎ • process returns with SLA tracking and 63 more.... When @shopify AI Toolkit was announced yesterday, we already knew what is the first thing we gonna do with it. Comment WOW and get the library link before everyone else. DM for contributions. Served with ❤️‍🔥 by 40RTY.

@forgebitz Clean codebase makes AI output better. How much does context window size matter?

code quality now matters even more: ai is really good at copying if you have a very good, clean, codebase to start with, ai output will be better if you start with slop it expands that

@n_kun_log Atlassian turning Confluence into an AI agent playground means documentation finally thinks back

📊 Atlassian just turned Confluence into an AI playground Visual AI tools + third-party agents = your wiki now thinks for itself Finally, documentation that writes back. No more dead knowledge bases. Which agent would you connect first? #AI #Enterprise

@garrytan Agentic Engineering day 2 with GBrain file storage improvements. This is the real build log

Day 2 of Agentic Engineering GBrain from the poolside of the Rosewood Kona Village Hawai’i Right now added improved GBrain file storage so that private Supabase Files linked from markdown can be clicked through via Supabase Edge Function temp redirects All from Telegram chat

@jainsahil Auth and memory breaking first in vibe coded apps is not surprising at all

Ran a poll asking what breaks first when a vibe coded app gets its first 100 real users. Auth, memory, and founder confidence tied at the top. Payments came last. Funny part: the code and the founder usually break together. What broke first for you?

@allenwalton Shopify inventory forecasting app built with Claude plus n8n workflows. Which nodes did you use?

Some things I've done with Claude lately •⁠ Built a Shopify inventory forecasting app •⁠ Setup n8n and built workflows I never would have been able to do previously •⁠ Analyzed ad campaigns across Meta and Google •⁠ Created PRDs for physical products I’m manufacturing •⁠ Moved personal site from Ghost Pro to my little Hetzner server •⁠ Cleaned up my Klaviyo account via API •⁠ Run reports on Stripe Billing •⁠ Built a half dozen skills to simplify workflows I’m using often •⁠ Migrated Shopify themes within just a couple of hours • Created an Airtable to manage my housefire rebuild estimates (insurance vs contractor) • Another Airtable for logging all personal property receipts, finances, and more •

@EXM7777 Jumping on trending AI models is quick money. Leveraging specific expertise with AI is wealth

best way to make quick money in AI is jumping on every trending business model... without a doubt best way to get filthy rich is leveraging your very specific expertise with an AI system that nobody else in the world can build and the second one is what everyone is sleeping on your edge doesn't have to be impressive... it could be experience from a specific industry, a small side project you did while studying, anything where you got decent results if you did something once and it worked, you can 100x it with AI now... for yourself and for everyone else who needs that exact thing one question remains: what different are you bringing to the table?

@pierreeliottlal Claude MCP pulling high intent LinkedIn leads and personalizing outreach is a real edge

This is how to use Claude MCP to get customers on LinkedIn. Ask Claude for high-intent leads Reach out with personalized messages based on real signals Turn replies into conversations Turn conversations into revenue

@n_kun_log Poke making AI agents as simple as texting is exactly what mass adoption looks like

🛠️ Poke just made AI agents stupid simple No more complex workflows or API calls. Just text and go. This is what mass adoption looks like - when your grandma can use AI agents without reading docs. But will simplicity kill customization? #AI #NoCode

@fivosaresti Moving clients from Claude Projects to Claude Code because Code actually executes. What stack?

We moved every client from Claude Projects to Claude Code. Projects was good for research and drafting. But CC actually executes. It: 1. Builds lists 2. Enriches contacts 3. Writes copy 4. Creates campaigns. Every client now has a living GitHub file: - ICP - Offer - Messaging - Tone - Case studies. MCPs pulling data automatically are connected to: - Findymail - Instantly - Apollo - HeyReach Plus, the file stays current without anyone maintaining it: - n8n syncs call transcripts. - Slack MCP reads client conversations. Build once, reused across the entire team. This is the way.

@OneJKMolina Skill issues becoming prompting issues assumes the prompts are being written well first

Very soon all skill issues in business will simply become prompting issues.

@iamcamengland AI agent at 5 minutes always wins over a salesperson responding in 4 hours

An AI agent that responds to your leads in 5 minutes will outperform a salesperson who responds in 4 hours every single time. Not because the agent is smarter. Because speed wins, and no human can be available 24/7 the way an agent can. My booking rate went from 15% to 28% just from that one change.

@lucasmeijer @_lopopolo Shipping a demo is easy. Real data with real users in workflows is the bar

Closing thoughts on ai engineer conf: - weather so great - did not like the venue one bit - can’t say any talks really stood out for me - but had very interesting hallway conversations. Especially with @_lopopolo who works opposite of me: he just lets it rip and not worry. 1/n

@marvinvista @_lopopolo @swyx @vibhuuuus Scarce resource in agentic builds is synchronous human attention not model capability

The fundamentally scarce thing now is the synchronous human attention of a team. When building with agents, start asking: - where is the agent making mistakes? - where am I spending my time? - how do I stop spending that time? h/t @_lopopolo, @swyx, and @vibhuuuus

@ChereneAubert Waiting until tech is easy means you always arrive after the edge is gone

The worst mindset you can have when it comes to emerging tech/AI is “I’ll wait until it’s easier”. It always becomes easy enough for the masses. Adopting before the masses do is the goal.

@georgeclem Building AI workflows when your constraint is leadflow is still the wrong order of operations

If your constraint is leadflow, playing with AI tools is a waste of time. My timeline is full of people building elaborate 'time saving' AI workflows. But these are usually the same people that struggle to book more than 3-5 sales calls per week. Saving time is not your issue.

@AdTurboAI AI handling execution while humans focus on strategy is finally the right split

What happens when AI handles execution: Humans focus on strategy. Humans focus on creative. Humans focus on business development. AI handles the tedium. Everyone wins.

@n_kun_log AWS betting on both OpenAI and Anthropic is the Switzerland play and it works

💰 AWS betting billions on both OpenAI AND Anthropic Their logic: "We're Switzerland, not picking sides" Smart move tbh. When you're the infrastructure, you win regardless of which model dominates. Is this the new playbook for cloud giants? #AI #Cloud