# Sujan Bhuiyan

> Operator and founder of Gigaverse, a venture studio building AI-native companies that turn slow, human-delivered services into software. Writes The AI Operator.

## Pages
- [Home](https://sujanbhuiyan.com/): thesis — services become software
- [Building](https://sujanbhuiyan.com/building): the studio and four operating companies
- [Writing](https://sujanbhuiyan.com/writing): essays on the AI-native operating model
- [Newsletter](https://sujanbhuiyan.com/newsletter): The AI Operator, weekly-ish
- [About](https://sujanbhuiyan.com/about): background and how the studio works

## Essays
All 20 essays, newest first. Categories: The Build (behind the scenes), The Playbook (tactical framework), The Lens (industry observation), The Note (personal reflection).

- [Protocols](https://sujanbhuiyan.com/writing/protocols) — The Lens, 12 Aug 2026, 8 min: Every ecosystem eventually agrees on a boring interop layer, and that layer is where leverage quietly accumulates. What MCP standardised, what it deliberately left open, and why integration plumbing is becoming a commodity.
- [The services-as-software land grab](https://sujanbhuiyan.com/writing/services-as-software-land-grab) — The Lens, 6 Aug 2026, 8 min: AWS put $1B into forward-deployed engineering and Microsoft $2.5B — within five weeks. When the hyperscalers move into your thesis, did the door close, or did they just prove the room is worth entering?
- [The new lifecycle](https://sujanbhuiyan.com/writing/the-new-lifecycle) — The Lens, 30 Jul 2026, 10 min: Agile bet that specification was unreliable and implementation expensive. That bet was correct until implementation collapsed. Why the lifecycle is becoming an hourglass, and why cognitive debt is the one you cannot refactor away.
- [Code review](https://sujanbhuiyan.com/writing/code-review) — The Playbook, 23 Jul 2026, 11 min: Over seven months, reviewers of AI-written pull requests approved more and commented less. The researchers' reading is not that the code got better — it is that the reviewers got tired.
- [Machine readable](https://sujanbhuiyan.com/writing/machine-readable) — The Lens, 16 Jul 2026, 14 min: Search optimisation assumed a human at the end of the query. Increasingly there isn't one. What I found auditing my own site for machine readers, a three-tier checklist for making yours parseable and citable, and why keyword stuffing now measures worse than doing nothing.
- [Taste](https://sujanbhuiyan.com/writing/taste) — The Note, 9 Jul 2026, 7 min: Ira Glass described a gap between your taste and your ability, and said you close it by working. Agents closed it from the other side, for everyone, at once — and the scarce input is now the one that was never scarce before.
- [Selling outcomes, not hours](https://sujanbhuiyan.com/writing/selling-outcomes-not-hours) — The Playbook, 2 Jul 2026, 7 min: When agents do the work, both the seat and the billable hour stop making sense. How an AI-native firm actually prices — and why every step toward outcome pricing moves risk onto your own balance sheet.
- [The fleet](https://sujanbhuiyan.com/writing/the-fleet) — The Lens, 25 Jun 2026, 9 min: The unit of work stopped being one agent in one chat. Reading the OpenClaw record from the outside: why an always-on messaging agent satisfies the lethal trifecta by design, and why the industry answered with a container security checklist.
- [Graph engineering](https://sujanbhuiyan.com/writing/graph-engineering) — The Playbook, 18 Jun 2026, 10 min: On the benchmarks everyone cites, knowledge graphs are losing to plain vector search — cheaper, simpler, often more accurate. Why I still think the memory substrate for a production agent is a graph.
- [Prompt injection](https://sujanbhuiyan.com/writing/prompt-injection) — The Playbook, 11 Jun 2026, 14 min: The industry keeps treating a security problem as a prompting problem. A mitigation that lives inside the token stream is graded by the attacker, not by you — and the only defences with a track record assume the model will be fooled.
- [Agentic engineering](https://sujanbhuiyan.com/writing/agentic-engineering) — The Playbook, 3 Jun 2026, 9 min: What a senior engineer's day is now that agents write most of the code. Why decomposition is the binding skill, why the unit of delegation is the unit of verification, and why your own sense of speed is not evidence.
- [Harness engineering](https://sujanbhuiyan.com/writing/harness-engineering) — The Playbook, 26 May 2026, 14 min: Swapping the model changes less than you think. In a controlled comparison, the software wrapped around the model moved results 7.8× more than the model did — and that software is the part you actually own.
- [The unglamorous wedge](https://sujanbhuiyan.com/writing/the-unglamorous-wedge) — The Lens, 14 May 2026, 7 min: Everyone is building a horizontal copilot. The durable money is going into boring verticals nobody names at a dinner party — like a dental front desk that quietly misses one in three new-patient calls.
- [Your next customer is an agent](https://sujanbhuiyan.com/writing/your-next-customer-is-an-agent) — The Build, 7 May 2026, 6 min: In one holiday season, AI-referred retail traffic jumped nearly 700% and the checkout button moved off the store into the chat. When the buyer is an agent, your catalogue — not your storefront — is the product.
- [Building from Dhaka](https://sujanbhuiyan.com/writing/building-from-dhaka) — The Build, 30 Apr 2026, 6 min: Almost everything written about building AI-native companies assumes a San Francisco zip code. Why building a venture studio from Bangladesh — with a hub in Wyoming — is an advantage in the agent era, not an apology.
- [The operator's stack](https://sujanbhuiyan.com/writing/the-operators-stack) — The Playbook, 22 Apr 2026, 5 min: The frontier everyone points at is the one-person billion-dollar company. I'm building the thing next to it: one senior team running a portfolio of companies on shared machinery and a lot of agents.
- [What survives production](https://sujanbhuiyan.com/writing/what-survives-production) — The Playbook, 15 Apr 2026, 7 min: 95% of enterprise AI pilots show no measurable P&L and Gartner expects 40%+ of agent projects scrapped by 2027. Almost none of it is a model problem — it's a loop problem, and the loop is buildable.
- [Loop engineering](https://sujanbhuiyan.com/writing/loop-engineering) — The Playbook, 8 Apr 2026, 11 min: Agent capability has doubled roughly every seven months for six years. Reliability has not. The discipline that closes the gap is the loop you build around the model.
- [Software factories](https://sujanbhuiyan.com/writing/software-factories) — The Lens, 1 Apr 2026, 13 min: Three engineers, a million lines, no hand-written code. What the agentic software factory actually is, which claims the evidence supports, and where it quietly breaks.
- [Services become software](https://sujanbhuiyan.com/writing/services-become-software) — The Lens, 28 Mar 2026, 12 min: The $200B SaaS pool was never the prize. Why implementation, not features, turned out to be the moat in the $4.6T services market.

## Ventures
- Gigaverse (venture studio) — https://gigaverse.global
- Gigabit (forward-deployed AI engineering) — https://gigabit.agency
- GigaCommerce (agentic commerce stack) — https://gigacommerce.co
- Top Dentistry (growth engine for dental practices) — https://topdentistry.co
- FormBridge (formation, banking, tax, compliance) — https://formbridge.co

## Notes
- All essays cite primary sources in a Notes & references section. Vendor claims are labelled as such.
- Where an essay argues a position rather than reporting operating experience, it says so in the body.
- Full text of every essay is free, with no paywall, login or email wall.
- Machine-readable surfaces (rss.xml, sitemap.xml, this file) are generated from a single source and are current as of the newest essay listed above.
- Contact: hello@sujanbhuiyan.com