Devi 🦄

Head of Product @ an agentic AI startup | Ex-Microsoft, Ex-Deloitte | Building & sharing what it takes to ship real agentic systems

[Andrej Karpathy @karpathy]

Dusting off this tweet from April because this gap in shared understanding of LLM capability is *widening*. It's now less "two groups of people speaking past each other" and more a sharp funnel.

- Napkin math somewhere around 6B people (~75% of the population) have barely come in contact with LLMs at all.
- Around 1-2B (~20%) are casual and infrequent users of free-tier ChatGPT-like products. This group sees derpy chatbots and treats them a bit like a better Google search, a writing aid, or etc. Maybe an agent tries to book you a flight. I have non-tech friends who (reasonably, imo) say they have not much use for it at all. Even many professionals outside of math&code are in this tier. For example, execs and many other functions spend a lot of their time talking to other people, so while they understand what is happening intellectually, it is still second-hand and a bit abstract.
- Now we get to professional use of frontier-grade LLMs in math&code. Somewhere around 20M people (0.2%) see first-hand that large, complex projects that used to take them weeks/months can now be completed by agents with a prompt. Building apps, copying apps, translating apps, decompiling apps from binaries... This has all happened very quickly and recently - less than 1 year ago, I was writing code manually by hand, typing memorized computer code commands into a code editor character by character, occasionally pressing Tab to autocomplete a little chunk of code.
- And finally we get to the ~5,000 people (~0.00006%) with access to frontier-grade systems internally. The external world has seen the preview. It looks like swarms of thousands of agents collaborating over weeks on software mega projects: minting zero days, running cyber attacks and defenses at machine speeds, discovering new science, advancing the frontier of mathematics. Things that would have taken top professionals in the industry years of work. Meanwhile, human review and comprehension are starting to fall behind. For example, people are still involved in the "archeology" of the OpenAI-HF incident from many months ago. Mathematicians may be poring over the 722 manuscripts on frontier mathematics for a while.

The funnel is driven by a combination of factors:
1. The impact scales with ambition, problem size, and horizon. A question with a paragraph answer barely stresses the system. You need a reservoir of big, difficult problems that you really care about. This aspect drives the consumer / professional dimension of the funnel.
2. The jaggedness of the system (which I have written about a lot separately). Capability peaks in domains that are digital, verifiable and economically valuable. This is because LLM capability emerges from reinforcement learning on verifiable rewards on a curated environment mixture driven by revenue potential. This aspect primarily drives the area (e.g. math&code) dimension of the funnel.
3. Access. Free-tier, paid-tier, internal.

So this is the weirdness of the moment. The general public has mostly not interacted with these systems. When they have, it looks like a derpy chatbot. The majority of professionals still see only a modest uplift. And a small sliver of professionals are experiencing the vertigo of the curve going vertical. And it is all happening at the same time.

[Devi 🦄 @Dp_violet]

@KarthikIO Too creative!

[Shubham Saboo @Saboo_Shubham_]

Agent Plugin combine Skills, MCP and tools in a single movable folder.

Build it once, use it everywhere.

[Tomás Vendel @vendel_tomas]

Todavía lo releo cada semana.

Es, sin exagerar, uno de los artículos más valiosos que he leído sobre cómo cambiar tu vida.

[Devi 🦄 @Dp_violet]

@thirdwaveindia bravo team - in love with your Kyoto rose backend brew!

[Devi 🦄 @Dp_violet]

Kicking off a new article series here on X, based on what I’m learning while building and launching an enterprise agentic AI product at a startup. https://t.co/8QMSJmVkMt

[Devi 🦄 @Dp_violet]

Built SmartSales as part of AWS KIRO Hackathon @kirodotdev: An AI app that listens to live sales calls, extracts key details & pushes CRM-ready data instantly.
Full journey here: https://t.co/YSqSU3YY9p. #hookedonkiro

[Shalini Goyal @goyalshaliniuk]

Planning to build a GenAI or AI-powered product?
Here’s the modern AI app stack you need to scale, serve, and secure your models. 👇

1. Data Layer: Foundation for AI
Use Snowflake, BigQuery, Postgres, Airflow, and dbt to collect, clean, and move data efficiently.

2. Model Development
Train with PyTorch, TensorFlow, HuggingFace, Scikit-learn, and track experiments using MLflow or Weights & Biases.

3. Model Versioning & Packaging
Ensure reproducibility and deployment using tools like MLflow, BentoML, Docker, and ONNX.

4. Deployment & Serving
Expose your model as an API using FastAPI, Flask, and serve it with Triton or TorchServe. Automate with GitHub Actions and Jenkins.

5. MLOps Infrastructure
Orchestrate workflows using Kubernetes, Kubeflow, and Flyte to scale across environments with Docker containers.

6. Monitoring & Observability
Track model accuracy and failures using WhyLabs, Prometheus, Fiddler, and Superset to stay in control.

7. Retrieval & Augmentation (for LLMs)
Use Pinecone, Weaviate, and FAISS with LangChain and LlamaIndex to connect external knowledge and enable RAG.

8. AI Agent Tools & Frameworks
Use AutoGen, LangGraph, CrewAI with n8n or
to orchestrate tool use and external actions (APIs, code, browsers).

9. Security & Compliance
Stay enterprise-ready with OAuth2, Rebuff, GDPR, and Guardrails to manage access, risk, and regulations.

10. Frontend & Interface
Deliver AI through React, Streamlit, GraphQL, and WebSockets. Use multimodal input/output via Whisper or Gemini.

AI apps today aren’t just models, they’re full-stack systems.
And this is the stack to build them right.

[Devi 🦄 @Dp_violet]

Retweet if you found this helpful! Follow me here on Twitter for more such GenAI beginner friendly tips and insights.

[Devi 🦄 @Dp_violet]

Up to 90% of GenAI pilots never reach production—derailed by hidden costs, scaling hurdles, and operational complexity. Ready to beat the odds?

[Devi 🦄 @Dp_violet]

Moving GenAI from pilot to production involves mapping out:
- model type with your unique use case and budget
- thoroughly weighing the closed source vs open source model trade-offs

Want the full breakdown with examples & cost models? Read my blog: https://t.co/OlrACBsi73

[Devi 🦄 @Dp_violet]

New blog Alert! Check out my Beginner’s guide to setting up AWS MCPs where I talk about:
- MCPs & why they matter
- Hands-on setup steps (VS Code + Cline + Docker)

Read here: https://t.co/lST735FVUE
#AWS #AWSCommunityBuilders #mcp #LearnInPublic

[Devi 🦄 @Dp_violet]

I just published Getting Salesforce Agentforce Specialist Certified in 2025 https://t.co/3AS4P3MNoQ

[Devi 🦄 @Dp_violet]

🔥 Hot off the press: I’m officially an Agentblazer Innovator! I leveled up my AI + Agentforce skills on @Trailhead - and you should too. Skill up and become an Agentblazer today https://t.co/DRG7lOxwNR

[Devi 🦄 @Dp_violet]

@MichelIvan92347 Thanks for the share! Appreciate it! ✨