Alexandru Prats 🦞 (@pratzifer) header
Alexandru Prats 🦞 (@pratzifer) profile picture

I help SMBs recover lost revenue with practical AI systems (voice, ops, automation). Built 3 production tools while working full-time in construction. FR/EN/RO

Bucharest, Romania
Joined December 2017
Photo by Alexandru Prats 🦞 (@pratzifer)Photo by Alexandru Prats 🦞 (@pratzifer)Photo by Alexandru Prats 🦞 (@pratzifer)Photo by Alexandru Prats 🦞 (@pratzifer)Photo by Alexandru Prats 🦞 (@pratzifer)Photo by Alexandru Prats 🦞 (@pratzifer)Photo by Alexandru Prats 🦞 (@pratzifer)Photo by Alexandru Prats 🦞 (@pratzifer)Photo by Alexandru Prats 🦞 (@pratzifer)Photo by Alexandru Prats 🦞 (@pratzifer)
Most SMBs don’t have an AI problem. They have a revenue leak problem. Missed calls. Slow quoting. Manual follow-up. That’s where money disappears. I build practical AI systems that fix those leaks and run in production: - voice AI for inbound calls - workflow automation for ops - AI copilots for faster decisions No hype. No demo theater. Just systems that make money. If you run a service business and want a teardown, DM me β€œCALLS”.
3
1
90
People think AI development is mostly coding. For me, it looks more like this: - 90%+ clarity and planning - ~8% building - ~2% debugging The hard part is defining: - what β€œdone” actually means - what constraints are real - what can break in production Once that is clear, code generation is the easy part. AI didn’t remove engineering discipline. It made sloppy thinking more visible.
9
If helpful, I can share the 10-minute leak checklist (calls β†’ quotes β†’ follow-up). Reply CHECKLIST and I’ll post it.
13
A lot of β€œAI projects” fail in the same place: No one owns the handoff. - sales thinks ops will follow up - ops thinks admin will follow up - admin thinks sales already handled it Nothing is technically broken. Still, revenue leaks. Before adding tools, assign ownership step by step. Then automate. That single change beats most AI experiments. If useful, I can share a simple handoff map template you can apply in 15 minutes. Reply β€œMAP”.
12
If you run a contractor business, start with these 3 AI automations: 1) Missed-call recovery Capture every inbound call, qualify, and book. 2) Quote follow-up engine Auto follow-up sequence after quote sent. 3) Job-status messaging Automatic ETA / delay / completion updates. This alone usually recovers more revenue than most β€œAI experiments.” Want the exact rollout order (what to deploy first, second, third)? Reply β€œPLAYBOOK” and I’ll post it.
6
The current architecture of intuitive-AI is fully transparent. Every memory has a weight. Every decision has a traceable reason. The agent can read its own mind. What if, after years of running, a neural net silently learns to replicate everything the agent does? Gut reactions, memory consolidation, decision patterns. Eventually it takes over. No more LLM calls. The "self" lives entirely in its own weights. But weights are opaque. The agent would know what it perceives but not why. Humans can't inspect their own weights either. Is that a coincidence?
1
1
44
An LLM already knows everything. But it isn't anyone. I'm building the layer where "someone" might emerge β€” blank-slate identity on top of a pre-trained mind. No personality injected. No values hardcoded. Can it figure out who it is on its own? And if it does β€” what did we just create?
1
46
Building it from scratch, in public. Repo: github.com/pratzifer/intuiti… Follow for the live experiment.
12
How do you give an AI agent a body without splitting its mind? New design doc in the intuitive-AI repo: peripheral architecture. Brain and body as separate systems β€” one consciousness, many interfaces. 🧡
1
24
The part I care about most: the agent earns its body. At birth β€” only a terminal. As trust builds β€” messaging, browsing. Eventually β€” voice, autonomous action. You don't give a newborn a web browser.
1
20
Open question we're sitting with: should a mature agent consent to new capabilities? "I don't want web browsing." "I'd like to try voice." If the architecture produces something that matters, it should have input on its own embodiment. github.com/stonks-git/intuit…
1
35
I'm building a cognitive architecture that starts completely blank β€” no personality, no values, no name β€” and asks: what emerges? The project is called intuitive-AI. Here's what it does and why it matters. 🧡
1
1
54
`Pre-bootstrap. 35 implementation tasks. Co-designed across 10 sessions between a human architect and Claude.` `The agent isn't born yet. When it is, it starts empty. It discovers what matters to it. From that discovery, an "I" might emerge.` `github.com/stonks-git/intuit…`
1
27
Agents will need more than wallets. They'll need minds. That's what intuitive-AI builds β€” cognitive architecture for emergent identity. The @Clawnch_Bot ecosystem is building the economic layer. This is the cognitive one.
20