The network for creativity
Join 1.25M professional creatives like you
Connect with clients, get discovered, and run your business 100% commission-free
Creatives on Contra have earned over $150M and we are just getting started
Back to feedPost
This week I went down a rabbit hole after reading about a Netflix engineer who found that up to 90% of the tokens fed into AI models during agentic workflows are pure waste bloated logs, raw JSON, repeated file trees.
I checked my own setup. Turns out I was doing the same thing: pasting full API responses into Claude Code instead of trimming them, letting file trees balloon in context instead of scoping the read.
Small fix, immediate difference shorter agent runs, fewer stalls, faster iteration on client builds. It's a good reminder that in AI-assisted development, the skill isn't just prompting well. It's knowing what NOT to hand the model.
If you're building with AI tools daily, it's worth auditing your context the same way you'd audit a slow database query.
The network for creativity
Join 1.25M professional creatives like you
Connect with clients, get discovered, and run your business 100% commission-free
Creatives on Contra have earned over $150M and we are just getting started
Trending
retrofuturism
Share your favorite designs from the past exploring the distant future. Or your own take on the aesthetic 🚀
halftone
Dots. Texture. Grain. Let's see your designs using the halftone technique.
Midjourney
Creatives are showing off what they're making with Midjourney.
colorgrade
Video editors sound off: How do you enhance the mood of your video in color grading?
Related posts
🚀 Introducing BrandOps — Your AI-Native Personal Brand Operating System
What if your AI didn’t just respond to prompts — but actually understood your professional identity, remembered your goals, learned from your achievements, and helped turn them into meaningful action?
That’s what I’ve been building with BrandOps.
BrandOps creates a persistent Digital Twin from your expertise, professional history, goals, preferences, projects, and verified activity — then transforms natural-language conversations into structured plans and executable workflows.
The operating loop:
Ask → Plan → Approve → Execute → Verify → Learn
🧠 Persistent Digital Twin
🤖 Specialized AI agents
🎯 Context-aware planning & next-best actions
✅ Human-in-the-loop approvals
⚡ Executable workflows
📈 Verified activity & continuous learning
🔌 Designed for interoperability with AI and developer ecosystems, including MCP
The goal is bigger than generating another piece of content.
BrandOps is an intelligent operating layer for your professional identity — one that understands what you know, what you've built, where you're going, and what actions can move you forward.
It's designed especially for AI builders, developers, founders, creators, and knowledge workers who are constantly building but need a better way to transform that work into visibility, positioning, and opportunities.
🌐 Try BrandOps:
https://brand-ops-navy.vercel.app/
💻 Explore the GitHub repository:
https://github.com/MISHHF93/BrandOps
👤 See BrandOps in my Contra portfolio:
https://on.contra.com/kMPGzj
I'm building toward a future where AI doesn't just know your latest prompt.
It knows your professional trajectory — and helps you act on it.
#BrandOps #ArtificialIntelligence #AIAgents #DigitalTwin #PersonalBranding #BuildInPublic #OpenSource #MCP #FutureOfWork
The Ask, Plan, Approve, Execute, Verify, Learn loop is what stands out to me here, most "AI does your branding" tools skip straight to execute. Curious how much people actually end up tweaking at the approve step versus just rubber-stamping what it plans.
DayMate is a lightweight AI-powered daily planning assistant built with React, Node.js, Express and Groq AI. Users can add, complete and delete their daily tasks. When they click 'Plan My Day with AI', the React frontend sends the tasks to an Express backend through a POST API. The backend securely handles the Groq API key using environment variables and sends the user's tasks to an LLM with a structured prompt. The AI generates a short and realistic plan for the day, which the backend sends back to React and displays in the UI.
Love this AI daily planner concept! Using React, Node, and Groq together makes it super powerful
AI-Agent Reporting Automation
Production AI-agent service that turns stakeholder requests into verified BI reports end to end: a request posted in Basecamp triggers a gated agent workflow that validates SQL against approved business logic, creates or updates the Metabase report, and verifies every write by exact read-back before replying with evidence. Containerized Node.js services (Docker Compose) with a fully offline test suite and CI. Built to cut the analytics team's ad-hoc reporting load - and to prove AI agents can be trusted with production writes when properly gated.
Solid setup. One thing that has saved me a lot in write-capable agents: log every agent turn (prompt, tool calls, outcome) to an audit table and add a replay mode — when a report comes out wrong, you can reproduce the exact run instead of debugging blind. Also worth setting a...