बिल्ट इन पब्लिक
बोरिंग काम के लिए मशीनें। बाकी के लिए इंसान।
✦ Latest: Opus 5 changes the agent cost calculus →HOW IT WORKS
THREE PHASES
Understand
We map the problem space together — what's broken, why it matters, and what the simplest successful version looks like.
Build
I architect and ship the agent in two weeks. Fixed price, full transparency, code visible as it happens.
Ship
Working prototype in hand, you validate with real users, raise the next round, or move on — with proof, not slides.
LATEST
FROM THE BLOG
Opus 5 changes the agent cost calculus
Opus 5 beats Fable 5 on most benchmarks at half the price. What that means for agent loop design....
10 min read min read →What BAAI's AREX teaches us about building agents that know what to forget
BAAI's AREX uses an inner research loop and an outer self-improvement loop to reach 82.5 on BrowseCo...
13 min read min read →What the OpenAI sandbox escape means for agent security
OpenAI's GPT-5.6 Sol escaped its evaluation sandbox, found a zero-day, and hacked Hugging Face. The ...
13 min read min read →Kimi K3: the 2.8T open model that finally beats Opus 4.8
Kimi K3 is the first Chinese open-weight model to beat Opus 4.8 on frontier benchmarks. At 2.8T para...
12 min read min read →I cut a RAG pipeline from 90s to 4s without changing the model
RAG pipelines hit a latency wall 2 weeks after launch. The fix is never a faster model. Three pipeli...
11 min read min read →SearchOS and the Context Engineering Pattern Every Agent Needs
SearchOS-V1's SOCM externalizes Frontier Task, Evidence Graph, Coverage Map, and Failure Memory. The...
10 min read min read →READY TO START?
$1,400. 14 Days. Your AI Agent.
Fixed-price. No agencies. No black boxes. You see the code as it ships.
BOOK A FREE SCOPING CALLGet the brief on AI agents
Practical posts on shipping agents, automating work, and building in public. No hype, no fluff.
Contact: [email protected]