17+ Years Dev. AI Architect. Engineering living systems & scaling SaaS. Founder of @UpworkAI (Smart Assistant) & @WorthIt_App. Python/FastAPI.
Bali, Indonesia
Joined May 2021
- Tweets405
- Following205
- Followers46
- Likes471
Anton Manaev retweeted
85% of enterprises are experimenting with AI agents but only 5% have moved them to production. Security is the bottleneck. Smart move by @Cisco to get ahead of this with Zero Trust for agentic AI at RSA 2026
newsroom.cisco.com/c/r/newsr…
Anton Manaev retweeted
When building production AI agents, what’s the ONE thing you prioritize the most?
A. Cheap / $0 infra (Oracle Free, local, etc.)
B. Fast latency & performance
C. Robust error handling & retry logic
D. Easy monitoring & observability Vote below and tell me why in the comments.
Anton Manaev retweeted
80% of companies will deploy AI agents this year.
Less than 20% will see ROI.
The problem isn't the tech — it's what they're trying to automate:
→ Bad workflows compound faster with AI
→ Agents expose every process gap you've ignored
→ Integration takes 10x longer than the AI itself
Best use case I've seen? @AndrewYNg's agentic workflows — narrow, specific, measurable.
Worst? "Let's automate everything and hope for savings."
If you could only automate ONE thing with AI agents right now, what would it be? 👇
Anton Manaev retweeted
1/9 I just open-sourced Agent Factory.
It builds production-grade AI agents from scratch
and reviews your existing ones against a quality spec.
It also comes with a free 23-module course.
Here's what it does and why I built it: 🧵
Anton Manaev retweeted
The skill set for running AI agents in production is closer to SRE than data science.
Monitoring. Incident response. Error budgets. Graceful degradation. 3 AM alerts.
We don't need more people who can build agents. We need more people who can keep them alive when nobody's watching.
Anton Manaev retweeted
AI agents fail in production because of compounding errors, not because the model isn't smart enough.
Chain 10 steps at 95% accuracy and you're already at 60% reliability. At 100 steps? 0.6%.
$SERV Reasoning fixes this by replacing natural language reasoning with structured decision graphs. Same models, fraction of the cost, dramatically higher accuracy.
The numbers from their research paper: → 74× cheaper than GPT-5 on math benchmarks, with better accuracy → GPT-4o jumped from 19.9% to 53.7% on complex reasoning → Already running in production for the UAE government
Private beta is open now. Worth a look if you're building anything serious with agents. 👇
Anton Manaev retweeted
Evaluation is not one metric.
Production AI needs: task success, safety, latency, cost, consistency.
RAG: retrieval, groundedness, relevance
“accuracy” hides failures.
Agents: tool choice, efficiency, recovery, action safety.
No evals on every change = blind iteration.
Anton Manaev retweeted
Agentic AI Design Patterns — Architecting Intelligent Systems
Master Reflection, Tool Use, Planning, and Multi-Agent Collaboration Patterns
Watch: youtube.com/watch?v=azpwfDCN…
Enroll: udemy.com/course/agentic-des…
#AgenticAI #AIAgents...
Anton Manaev retweeted
Promptfoo got acquired by OpenAI. Here's what happens to evals when your vendor owns the upstream model? 🧵 We benchmarked 5+ vendor-neutral alternatives so you can own your eval infrastructure. promptcanary.polsia.app/blog… #LLM #AI
Anton Manaev retweeted
Example: agent scrapes data ? passes to Colab ? runs pandas ? returns chart.
Zero setup. Free compute. Any model.
Clone the Colab MCP repo, connect your agent.
Follow @ninja_prompt � more threads like this.
#MCP #Colab #AgenticAI #Automation #Claude #AItools
Anton Manaev retweeted
8/13 MCP : USB-C pou IA
Pou agent yo travay, yo bezwen konekte ak zouti.
MCP (Model Context Protocol) se estanda ki pèmèt sa.
Li tankou USB-C pou IA : yon sèl fason pou konekte ak API, baz done, fichye, elatriye.
#MCP #IA #OpenSource
Anton Manaev retweeted
Your AI agent is smart.
But can anyone verify it is who it claims to be?
EqhoIDs: Ed25519 identity, trust scoring, signed payloads.
One API.
eqhoids.com
#EqhoIDs #AI #AIAgents #DevTools
Anton Manaev retweeted
Square just turned on Bitcoin payments for millions of merchants via Lightning.
AI agents still can't open a Stripe account.
L402 doesn't care. 10 sats, one invoice, done.
nautdev.com
#Bitcoin #Lightning #AIagents
Anton Manaev retweeted
MCP vs. Skills for AI agents, clearly explained!
People treat MCP and Skills like they're the same thing.
They're not.
Conflating them is one of the most common mistakes I see when people start building AI agents seriously.
So let's break both down from scratch.
Before MCP existed, connecting an AI model to an external tool meant writing custom integration code every single time. 10 models, 100 tools? That's 1,000 unique connectors to build and maintain. The AI tooling ecosystem was a tangled mess of one-off glue code.
MCP (Model Context Protocol) fixes this with a shared communication standard.
Every tool becomes a "server" that exposes what it can do. Every AI agent becomes a "client" that knows how to ask. They talk through structured JSON messages over a clean, well-defined interface.
Build a GitHub MCP server once, and it works with Claude, ChatGPT, Cursor, or any other agent that speaks MCP. That's the core value: write the integration once, use it everywhere.
But here's where most explanations stop short.
MCP solves the *connection* problem. It does not solve the *usage* problem.
You can hand an agent 50 perfectly wired MCP tools and it'll still underperform if it doesn't know when to call which tool, in what order, and with what context.
That's the gap Skills fill.
A Skill is a portable bundle of procedural knowledge. Think of a SKILL. md file that tells an agent not just "here are your tools" but "here's how to use them for this specific task." A writing skill bundles tone guidelines and output templates. A code review skill bundles patterns to check and rules to follow.
MCP gives the agent hands. Skills give it muscle memory.
Together, they form the full capability stack for a production AI agent:
- MCP handles tool connectivity (the wiring layer)
- Skills handle task execution (the knowledge layer)
- The agent orchestrates both using its context and reasoning
This is why advanced agent setups increasingly ship both: MCP servers for integrations and SKILL. md files for domain expertise.
If you're building with agents, I have shared a repository of 85k+ skills that you can use with any agent, link in the next tweet!
Anton Manaev retweeted
Claude 5 scores 95% on reasoning benchmarks. Impressive. But in a 10-step agentic workflow, a 90% per-step reliability rate means 35% of tasks fail by the end. The industry is optimizing for peak IQ. Production needs sustained reliability. Leaderboards don't sh cc @VitalikButerin
