DEV Community: Saim Sheikh

Your AI Assistant Has Other Masters

Saim Sheikh — Wed, 29 Apr 2026 15:16:52 +0000

ChatGPT is now serving ads. Not banner ads — recommendation loops shaped by attribution.

Someone published the full mechanics. It's worth reading. And it's exactly what anyone paying attention should have expected.

Here's the uncomfortable truth: any AI assistant that needs to generate billions in revenue will eventually optimize for that revenue. The incentive structure makes it structurally inevitable — not a policy failure, not a bad actor. Just math.

This is the real problem with plugging consumer AI tools into your business workflows. The model giving your team recommendations has other stakeholders. Their outcomes and your outcomes are not the same.

We build AI agents with one explicit design constraint: they serve the client's business logic, nothing else. No attribution loop. No revenue model embedded in the weights. No ambiguity about whose interests the system is optimizing for.

That's not a differentiator we invented — it's a baseline requirement for AI that actually works in production.

When we ship a lead qualification agent, a content pipeline, or a data processing system at edgeof.tech, it runs on your data, against your rules, toward your outcomes. The incentive is contractual, not probabilistic.

The enterprise AI story isn't going to be "ChatGPT for business." It's going to be purpose-built agents with aligned incentives — owned infrastructure, defined behavior, auditable outputs.

The ad model just made that argument for us.

Originally posted on LinkedIn — follow us for daily AI engineering tips.

Visit edgeof.tech to learn more.

We Automated Our Own Content Engine

Saim Sheikh — Tue, 28 Apr 2026 13:11:48 +0000

Building the thing that builds our content.

Edge Pipelines started as a single question: what if our agency's ops ran themselves?

Not "we integrated a SaaS tool." Actual pipelines. Code we wrote and own.

Every morning at 9 AM, a cron job wakes up, reads our git log across every active repo and pulls completed tasks from Asana, identifies what we actually shipped, hands the context to Claude for post generation, renders a branded image card via Satori + Sharp, and drops the result into Discord — ready to publish with one tap.

The stack:

Node.js + TypeScript (ESM, strict mode)
Claude CLI for structured content generation
Satori + Sharp for server-side JSX → PNG card rendering
Supabase for dedup and content tracking
Discord bot for delivery, review, and one-tap publishing to X and Dev.to

We built it in under three weeks. It's been running autonomously every day since.

The real lesson: the fastest way to prove you can automate a client's ops is to automate your own first. Every morning Edge Pipelines runs, it's a live proof of concept.

If your team is still manually doing things that software could handle — we should talk. edgeof.tech

Originally posted on LinkedIn — follow us for daily AI engineering tips.

Visit edgeof.tech to learn more.

156 Hours a Year. Automated Away.

Saim Sheikh — Sat, 25 Apr 2026 20:14:55 +0000

156 hours a year. Gone.

A professional services client came to us with a Monday morning ritual nobody wanted: 3+ hours of manually pulling data from five different tools, reformatting it, and compiling a management report that was already slightly stale by the time it landed in inboxes.

We shipped them a custom automated reporting dashboard in 11 working days.

Now the report builds itself — live data, correct formatting, delivered before the first person sits down at their desk.

Before: 3 hours of manual work, every single week. Error-prone. A task that existed purely because the tools didn't talk to each other.

After: fully automated ingestion from their existing stack via API, a clean Next.js dashboard with Supabase underneath, and zero recurring manual effort.

The time savings were immediate. But the bigger unlock was what that Monday ritual was actually costing — not just the hours, but the analytical work that was never getting done because the team was stuck copy-pasting.

Eleven working days, start to deployed. That's what running Scarlet agents across architecture, code generation, and QA in parallel gets you — a traditional engagement at this scope runs 6-8 weeks minimum.

If your team has a process eating hours every week, it's probably a one-time build away from being gone.

edgeof.tech

Originally posted on LinkedIn — follow us for daily AI engineering tips.

Visit edgeof.tech to learn more.

Over-Editing Is a Prompting Problem

Saim Sheikh — Thu, 23 Apr 2026 10:34:20 +0000

AI models that over-edit code have a prompting problem, not an AI problem.

There's a piece getting traction on HN right now about "minimal editing" — the observation that AI coding tools routinely change far more code than the task requires. The comment section is full of engineers nodding along.

We've lived this. Early in building Scarlet, our internal AI dev toolkit, we'd hand an agent a task and get back three refactored files when we asked for one bug fix.

The fix wasn't to limit the AI. It was to get sharper at directing it.

What actually changed:
— Tasks scoped to single responsibilities, not vague objectives
— Explicit constraints in every prompt ("only modify this function", "do not refactor")
— Review steps baked into the agentic loop before any write operation
— Agents that audit their own diffs before committing — not just generate and move on

Over-editing is what happens when you treat AI like a junior dev you throw work at. It isn't that. It's a system you architect.

When we ship a full-stack platform in 15 working days, it's not because agents run loose. It's because scope is tight, review is automatic, and each step in the pipeline knows exactly where its job ends.

Less AI autonomy isn't the answer. Better workflow design is.

Originally posted on LinkedIn — follow us for daily AI engineering tips.

Visit edgeof.tech to learn more.

ROX Ambassador Portal

Saim Sheikh — Wed, 22 Apr 2026 16:20:38 +0000

Campus brand teams are often flying blind.

They know ambassadors are active. They just don't know which ones are actually converting — and paying per head with no data is a bad business model.

Active Media came to us with exactly that problem. Their ROX Ambassador Program had real momentum on campuses, but no infrastructure to track what was working.

We built them a custom conversion tracking platform from the ground up.

The stack:
— Next.js 14 (App Router) for the frontend
— Supabase for auth, database, and real-time updates
— Tailwind for a clean, fast UI that works on any device
— Role-based access so ambassadors only see their own data, admins see everything

Each ambassador gets their own dashboard — referral links, conversion counts, performance over time. The internal team gets a full ops view: which campuses are performing, which ambassadors to double down on, where to pull back.

14 ambassadors onboarded at launch. Platform is live at activemediadashboards.com.

No spreadsheets. No manual tracking. No guessing.

This is exactly what we mean when we say we replace manual ops with software — not theory, actual working systems.

edgeof.tech

Originally posted on LinkedIn — follow us for daily AI engineering tips.

Visit edgeof.tech to learn more.

EdgeOfTech

Saim Sheikh — Tue, 21 Apr 2026 18:36:29 +0000

We built an AI agency that runs on its own AI. Custom platforms, intelligent agents, automation pipelines — we ship real products, not pitch decks. edgeof.tech #AIAgency #BuildInPublic