
Founder & CEO, NextCTL | AI agents for business + AI automation @theNextctl | ex-ResponseCRM | Building software since 2010
Dhaka · Serving US/UK/UAE
Joined May 2011
- Tweets531
- Following968
- Followers260
- Likes226
Agents who post daily don't film daily.
30 min on Monday → 5 clips → 25 posts → scheduled once.
Content isn't a daily chore. It's a weekly system. 👇
4
I auctioned off my social media stack 🔨
Buffer schedules. Canva designs. SocialCTL plans, makes AND posts.
SOLD. From $19/mo → socialctl.io
#AITools #SocialMediaMarketing
19
Your AI agent is burning money. 🔥
It calls a giant LLM for decisions a reflex could make.
3 rules for faster, cheaper agents:
1. Keep context outside the model
2. A decision can call another decision
3. It must work with no LLM at all
#AIAgents #LLM #AIEngineering
1
24
Business doesn't wait for calm. 🐊
Our AI agents capture leads, reply in seconds and keep your CRM in sync while the chaos rolls on.
Chaos outside. Control within.
nextctl.com
#AIAgents #Automation #NextCTL
25
AI coding agents are fast. The hard part is trusting what they ship.
I built AgentNeuron, a free, open-source control center for Claude Code, Codex, Cursor and Copilot. v1.1 is out today.
github.com/TurjoAt/agentneur…
3
3
102
Usage and cost per session, model and branch, with a step-by-step replay.
Plus truth and security checks for CLAUDE.md, AGENTS.md and MCP configs, and agent briefs so a new session picks up where the last one stopped.
1
1
11
robots.txt → where crawlers may go
schema.org → what your things are
WebMCP → what agents can do
Three layers of machine-readability, 15 years apart.
Your site is findable. It is not yet usable.
1
1
99
Spent the week wiring WebMCP tools into a product that already runs a server-side MCP.
Same business logic. Two completely different exposure surfaces. Two different threat models.
1
85
AI referral traffic converts at 14.2%. Google organic sits at 2.8%.
Five times better, and most brands cannot tell you whether they appear in AI answers at all.
#GEO
1
52
SocialCTL is in copilot testing.
It plans the calendar, writes the caption, picks the slot, publishes. I approve.
This tweet went through it.
#buildinpublic
2
55
Analysis of 6,259 production agents found a 56.6% task success rate.
Close to a coin flip. And most teams have no idea which half they are in because nothing logs at the step level.
Ship the observability before the agent.
#AIagents
34
11 major AI models shipped in 20 days this month.
Independent verification takes longer than that. Every benchmark you read is describing a model that has already been replaced.
Pin your version. Test on your own workload. Ignore the leaderboard.…
18
August rankings put Claude Code first for long autonomous coding sessions. Codex CLI for PR-shaped work. Cursor for in-editor.
Pick one. Standardize it.
Five developers improvising five different agent setups is not a stack. It is five stacks.…
1
44
Why boring wins in AI specifically:
The AI layer changes every six weeks. Models improve, APIs shift, pricing moves. If your foundation is also experimental, you are debugging two moving layers at once. Nobody wins that fight.
Boring tech has answers. Every problem we hit has been solved publicly by someone else years ago. Every exotic framework problem becomes a research project on our payroll.
Failures are predictable. After sixteen years I know exactly how these tools break. Predictable failure is what lets you write fallbacks that actually work.
#SoftwareArchitecture #TechStack #ProductionSystems #AI
34
Want more builders and founders on my timeline.
If you're:
Building a SaaS
Shipping a side project
Going solo on something ambitious
Building in public and sharing the wins, failures, and lessons
I’d love to follow your journey.
1
33
Founders ask me what running AI actually costs. The honest answer: it depends on which line items you bothered to count.
#SaaSEconomics #AIInfrastructure
1
83


