Skip to content
arafat@ops:~/dashboard

Engage superintelligence mode

I run Claude Code and Codex in six separate terminal panes across two tabs with the command fleet up 6 2 claude,codex. Each has its own task and git worktree. This replay illustrates research, build, review, security, memory and documentation working in parallel. Research links context, review verifies three example checks, security records my nine disclosed CVEs, memory mounts my second brain, and documentation prepares the handoff. The build waits until review verifies the change, then releases its commit and all six branches merge into main. The example filenames and test counters explain the workflow. Press Enter or Engage superintelligence mode to open my dashboard. Any other key or tap shows the completed replay. Skip intro and Escape enter immediately.

arafat / I delegated. Keep up.

Booting. Click the button to enter now, or press any key to skip ahead.

Easin ArafatApplication Security Engineer at Startise

Current focus agentic AI systems

The future is agentic.I'm already running it.

Smart is common. Smart and secure is rare.
That's the part I do.

I'm an Application Security Engineer at Startise. I run AI coding agents in parallel, design the review gates that keep them honest, and break my own systems before anyone else gets the chance. 40+ projects shipped while others were still choosing a tool.

9 CVEs disclosed / 40+ projects shipped / AI agents at work every daySee the work

The operator behind the fleet.

Easin Arafat, Application Security Engineer at Startise
Easin Arafat working at a laptop beside a terminal display

Security sets the guardrails.
The fleet carries the work.

A finding is only done when the fix ships.
WHAT I DO$ ./about_me.sh --roleMY SKILLS$ ./skills.sh --viewCONTACT ME$ ./connect.sh --secure

$ ls ~/work

my answer to repetitive work

I got tired of waiting.
So I built it.

I want context that sticks, sessions that survive, and changes I can inspect.
So I built the tools.

I make the rules.
Intent
I set the direction.
Context
I keep useful knowledge close.
Action
I put the tools to work.
Review
I inspect the result.

I take security personally.Check my findings

I made these.

Check my GitHub

I make "done" prove it.

Yoke

Local policy gates for Claude Code and Codex: human-approved checks, independent review and a console that shows the decision trail. A pre-release tool built around proof of completion.

Source Docs

I gave multitasking a floor plan.

herdr-setup

One-command setup for running Claude Code, Codex and Agy side by side in herdr. The fleet command starts workers, gives them tasks and keeps their work visible; an interactive manual lets you plan the layout first.

Source Open project Docs

I gave my books a plus-one.

DeepRead

A private reading library with an AI tutor beside the passage: word meanings, explanations, offline natural voices, precise bookmarks and shared libraries. Readers bring their own Claude Code, Codex or API key.

Source Open project Docs

I insist on a second opinion.

App Harness

A Claude Code skill for supervised app-building loops. Isolated planning, generation, deterministic gates and adversarial evaluation exchange evidence through files, so the builder does not grade its own work.

Source Docs

I keep the whole team in view.

Multi-Agent Orchestration

A coordination framework for agent teams, Kanban tasks, inter-agent messaging and multi-stage workflows, with shared memory and a signed audit trail. A dashboard keeps team activity and workflow execution in view.

Source Docs

I can see what ran.

Agentic Terminal

A folder-scoped terminal agent where shell commands and natural-language requests share one prompt. Tool calls, file changes and approval steps stay visible, with hosted models, local Ollama or a Claude Code CLI connection.

Source Docs

I gave parallel work its own space.

wt-cli

A shell-native Git worktree manager for listing, creating and moving between workspaces by number or name. It keeps parallel agent work in separate directories without making the operator memorize paths.

Source Docs

I turned repetition into a skill.

Day-to-Day Skills

A personal toolkit of reusable Claude Code skills, including article and security-research workflows. Structured instructions and supporting templates turn repeated work into a repeatable process.

Source Docs

A closer look at my work

I made two things.
Then kept going.

I built TermStream for persistent coding sessions and Second Brain.Deck for connected knowledge. Here is what I wanted, what I built, and why.

I BUILT IT / HERE'S THE TOUR

TermStream

A secure browser and mobile control plane for persistent AI coding sessions on infrastructure the user owns.

Explore TermStream Open ts.arafatops.com
TermStream dashboard showing server selection, workspace controls, and agent launchers
I built this view. See the active server, choose a workspace, launch the right tool, and return to running work from one dashboard.
See everything I built

$ ls ~/proof

my claims have links

I keep receipts.
You can check.

I disclose responsibly / Patchstack

9 CVEs.
I said it
was broken.

I found broken access control, IDOR, and data exposure in WordPress plugins. I bring that same suspicion to every system I build.

cd security-research ->

I CO-AUTHORED / ELSEVIER ARRAY

Adaptive UI for Mobile Banking - Enhancing UX through Machine Learning

Elsevier ArrayQ1 journalImpact factor 4.5Open access

DOI: 10.1016/j.array.2026.100901

cd featured ->
  • The Daily StarThey wrote about me
  • URC 2021My team won globally
  • Top 1% TryHackMeI play to win

$ ls ~/write

my brain, in public

I think loudly.
In writing.

17 Medium articles. I write about AI, security, business, psychology, and whatever else refuses to leave my head.

More things I had to say

$ ls ~/etc

I leave the keyboard, too

I have range.
Apparently.

2021:

I helped take a rover
to a global title.

My team won URC 2021.
I led software and communication for MIST Mongol Barota at the University Rover Challenge.

I build communities, read, compete, and occasionally remember the world has an outside.

I build communities

Leadership
President of MIST Cyber Security Club (2023-2024)

MIST LEETCON 2023
Event Organizer for Bangladesh's first international cybersecurity conference, hosting 3,500+ participants.

I make security a team sport

CTF Competitions
CTF Player and Problem Solver. Ranked top 1% in TryHackMe.

Flaghunt 2023
Event Host for National CTF Competition in collaboration with CTFBD.

Cyber Drill 2023
Organizer for Financial Institute & Critical Information Infrastructures Cyber Drill.

Cyber Threats Seminar
Host for emerging cyber threats seminar with Armed Forces Division of Bangladesh.

I do go outside

Sports
Active participant in sports like Football, Cricket, Kickboxing, Swimming, Table Tennis, Badminton, etc.

Hobbies
Love Reading, Writing, and Travelling.

$ ls ~/skills

my skills need their own orbit

I needed more
than one planet.

78 skill modules. 9 domains. I gave them a universe.

Launch pad ready.

cd skills ->

Pick a side of me.

4 sides of me connected.

I have receipts for that.

  • I FIND THE CRACKS

    I find broken access control, IDOR and data exposure. I disclose responsibly through Patchstack.

    cd security-research ->
  • I SHIP MY IDEAS

    I have shipped 40+ projects. I prefer my ideas running in production.

    cd projects ->
  • I FOLLOW THE MONEY

    I write about business because code is only part of what makes the world work.

    cd business ->
  • I BUILD THE THING

    I build security tools, DevSecOps automation and web apps I actually want to use.

    cd projects ->

$ ./connect.sh --secure

Make it difficult.
I get bored.

Give me a stubborn security problem, a tool that should exist, or an idea with some nerve. I like having something to prove.

TALK TO ME$ ./connect.sh --secure