I build AI systems that replace entire teams. Automation, marketing, sales. Sharing what actually works.
United Kingdom
Joined December 2019
- Tweets166
- Following201
- Followers71
- Likes1.5K
bots have officially overtaken humans on the internet.
not a prediction. a report from this week.
AI traffic grew 8,000% in one year. automated systems now generate more internet activity than people do.
most businesses are still marketing to humans through human-operated channels. but the buyer is becoming a bot.
shopify just launched agentic storefronts. your products sell directly inside ChatGPT and Gemini now. visa is testing AI-initiated payments.
the customer journey is being automated from both ends.
companies still running manual marketing, manual outreach, manual customer service are competing against AI systems that work 24/7.
AI systems cost $400/month.
a 3-person team costs $15,000/month.
same output.
the gap between automated and manual is growing every week. the internet moved on. has your business?
anthropic accidentally leaked their next AI model. it's called claude mythos. fortune broke the story an hour ago.
it's bigger than opus. a completely new tier called "capybara." anthropic says it's "the most capable model we've ever built."
here's what the leaked documents actually reveal.
#ClaudeAI #AIAgents
the timing matters. bloomberg reported today that anthropic is considering an IPO as early as october 2026. wall street banks are already in discussions.
mythos is the product that justifies that valuation. if it performs as described, every enterprise AI deployment gets rewritten.
sent claude a text from my phone at 8am. by 8:07 my laptop had opened 4 apps, filled a client intake form, and drafted the onboarding doc. i never touched my keyboard.
here's what actually happened, and what it means if you run an agency.
what i'd automate next if you run an agency: weekly client reports. invoice reminders. proposal drafts from discovery call notes. each one is 15-30 min of human time that follows the same pattern every time.
the real cost math: a VA doing this takes 25 min per client at $15/hr. that's $6.25 per onboard. claude does it in 7 min for roughly $0.40 in tokens. at 20 new clients/month, that's $125 saved. small number. but it compounds across every repeatable task.
the part most agencies miss: this isn't about saving 7 minutes. it's about removing yourself from the process entirely. i was on a morning walk. the client got onboarded before i sat down at my desk.
what claude actually did in 7 minutes: opened gmail and found the intake email. extracted client name, business type, goals, budget. opened the onboarding template. filled every field. saved the doc in the client folder. drafted a welcome email with next steps.
the setup: claude cowork + dispatch. you set up a project folder on your desktop, then send tasks from your phone remotely. works on mac and windows.
one prompt: "run client onboarding for [name]. use the google docs template, pull details from their intake email."
Alin Dobrin retweeted
how to become an ai automation and workflow expert and earn 4–5 figures in 6 months
by the end, you’ll be able to:
— design end-to-end automated workflows for real businesses
— link ai to the tools companies already rely on (crm, email, docs)
— replace repetitive tasks with ai-driven systems
— charge clients $500 to $5k per month with real results
month 1: build a solid base
focus on:
— learn python basics (just functional, not advanced)
— understand apis (http, json, webhooks)
— pick one no-code tool (make, n8n, zapier) and master it
— read api docs to connect tools
— get basic at prompt engineering (inputs, outputs)
project: automate a small personal task using make or n8n.
month 2: connect ai to your workflows
focus on:
— openai/anthropic api basics (completions, structured prompts)
— embed ai inside workflows (beyond just chatting)
— tool calling (ai triggers actions)
— build step-by-step chains (trigger → ai decision → action)
— track costs (token usage, when ai is overkill)
project: build a workflow that automatically sorts and routes incoming emails.
month 3: create real business pipelines
focus on:
— lead generation automation (scraping, enrichment, outreach)
— ai-powered personalized outreach at scale
— crm automation (auto-updating tasks, logging calls)
— content pipelines (idea → draft → publish)
— meeting automation (transcribe → summarize → log)
project: build a complete lead generation and crm workflow—either for a real or mock business.
month 4: introduce ai agents
focus on:
— what ai agents really do (no hype)
— when to use agents vs simple automations
— how to route tasks, manage steps, and keep state
— human-in-the-loop checkpoints (manual review when needed)
— build agent reliability (logging, retries, fallbacks)
project: create a support agent that handles basic customer queries, escalates tough cases, and logs all interactions.
month 5: get it production-ready
focus on:
— how to deploy (self-hosted n8n, make teams, or python backend)
— set up logging and monitoring (know when it breaks)
— version your prompts (no random changes mid-project)
— basics of security (protect api keys, access control)
— handle rate limits and downtimes gracefully
— document everything so you can hand off to non-technical clients
project: take one of your month 3 or 4 workflows and make it fully ready for a client, with proper docs and monitoring.
month 6: choose your income path and go all in
pick one direction:
freelance builder (fastest money)
→ focus on 2–3 repeatable workflows (like lead gen or support bots)
→ outreach to smb owners, agencies
→ start with small projects, then scale to retainers
in-house engineer (steady role)
→ focus on internal ops use cases
→ connect ai to company tools (slack, hubspot, etc.)
→ prove time and cost savings
build an ai automation agency (scale it)
→ pick a niche (e.g. real estate, e-commerce)
→ create repeatable services and templates
→ hire or partner to scale
the key is practice—every single month. once you follow this, you won’t just be learning ai—you’ll be using it to earn.
keep an eye out.
cc: @DeRonin_
a client replaced their 3-person outbound team with Apollo AI sequences. response rates went from 4% to 11%. cost went from $12k/mo to $800. what's stopping you from doing the same?
Nvidia announced NemoClaw at GTC 2026. A security sandbox for AI agents in regulated industries like finance.
Financial services had the biggest attendance share at the whole conference. The money is clearly moving toward agentic AI.
Jensen Huang expects $1T in orders by 2027
Undersea cables in the Red Sea, crucial for global communications, have been severed, impacting 25% of data traffic between Asia and Europe. The cause remains unknown, sparking debates on the vulnerability of our internet infrastructure. #RedSeaCable #InternetSecurity
