I am Shiv — a backend / systems engineer who looked at modern AI agents and immediately started asking annoying questions like:
“Nice demo. But where are the permissions? Where are the receipts? Who approved the robot touching production? Why is this workflow held together by vibes, a shell script, and fear?” 🤖🧃💥
So now I build toward systems that are useful, observable, recoverable, permissioned, and boring in production — chaotic only in the README.
My serious side: Go backend engineering, distributed systems, payments-grade correctness, observability, production services, data flows, reliability, and secure system boundaries.
My silly side: I treat bugs like tiny dungeon monsters and architecture docs like treasure maps drawn by a caffeinated raccoon. 🦝🗺️✨
|
APIs, services, workers, data flows, idempotency, reconciliation, and boring reliability magic. |
Agents, local-first context, inference runtimes, tool execution, memory, permissions, and audit trails. |
Receipts, rollback, capabilities, sandboxing, least privilege, and proof that the system did the right thing. |
class Shiv:
default_mode = "systems goblin"
favorite_problem = "messy system boundary"
usual_response = "wait, what happens when this fails?"
special_move = "turn chaos into boxes, contracts, logs, and receipts"
weakness = "overthinking infrastructure"
rare_drop = "a README with too many emojis"
The quest board says enough for now.
Some doors are still locked. Some goblins are still negotiating with product scope. 🧌🔐
No generic scoreboard here.
My real telemetry is whether the system became safer, clearer, and harder to break. 🧯✨
| 🗡️ Primary weapon | Go, backend services, distributed systems, APIs, workers, production debugging |
| 🧪 Lab tools | Rust, Python, Linux, local inference, agents, runtime boundaries |
| 🧰 Utility belt | PostgreSQL, Redis, queues, gRPC, observability, traces, logs, metrics |
| 🧾 Trust items | Idempotency, reconciliation, audit trails, rollback, capability design, sandboxing |
| 🍪 Emergency snacks | Architecture diagrams, TODO lists, coffee, stubborn optimism, and one suspicious shell script |
I like systems where every monster has a name, every risk has a boundary, and every failure has a recovery path. 🧌🧪🧯
graph TD
A["😵 messy problem"] --> B["🔍 find hidden boundaries"]
B --> C["🧱 define contracts"]
C --> D["🛡️ add permissions + safety"]
D --> E["📜 add logs, traces, receipts"]
E --> F["🧪 test weird failure paths"]
F --> G["🚀 ship the tiny rocket"]
G --> H["🐛 bug appears wearing sunglasses"]
H --> B
- 🧾 Commit receipts for AI-generated code.
- 🛡️ Capability-based permissions instead of “just trust the agent bro.”
- 🧠 Local-first AI systems where users own their context and memory.
- 🧰 Backend correctness: retries, reconciliation, recovery, idempotency.
- 📡 Observability that tells the truth when everything is on fire.
- 🏗️ Architecture that survives production, not just whiteboards.
- 🧌 Tiny infrastructure goblins that become serious products.
- 🪤 Failure traps that catch bugs before users do.
- 🚪 Hidden boundaries that explain why a system feels cursed.
| 🚀 Startup Mode | Give me ambiguity, constraints, and a hard problem. I will turn it into a buildable system. |
| 🧯 Production Mode | Give me a messy backend, failure-prone workflow, or unclear system boundary. I will chase the gremlins. |
| 🤖 Agent Mode | Give me AI automation that feels too magical. I will add permissions, audit, rollback, and trust. |
| 🧠 Research Mode | Give me a strange infrastructure thesis. I will overthink it until it becomes an architecture. |
🔭 Working on: hidden infra quests + Draft
🌱 Learning: deeper Rust / systems / AI runtime boundaries
💬 Ask me about: Go, backend systems, observability, AI-agent safety, weird infra ideas
🤝 Collaborate on: developer tools, platform engineering, runtime systems, secure automation
⚡ Fun fact: I name bugs like RPG enemies so they become easier to defeatI do not want AI systems that feel like haunted autocomplete.
I want systems with boundaries.
I want agents with receipts.
I want developer tools that make code safer, not just faster.
I want local-first infrastructure where users keep control.
I want production systems that fail loudly, recover cleanly, and explain what happened.
Also, I want all of this with more emojis.
Bring me a hard infrastructure problem, a cursed developer workflow, an AI-agent reliability mess, or a production bug wearing a tiny hat.
I will not just implement it. I will turn it into a system. 🧃🚀✨
