rashidae

That’s on you. If a person has good grammar and knows how to articulate his idea, it doesn’t make him an AI.

Does it have a CLI like github? Does it follow GIT? I made an account and did not migrate my repos, as I'd like to test this before moving in... Thanks in advance!

It was refined by chatgpt. ;)

Interesting. Have you tested other LLMs or CLIs as a comparison? Curious which one you’re finding more reliable than Opus 4.5 through Claude Code.

As a mirror to real-world agent design: the limiting factor for general-purpose agents is the legibility of their environments, and the strength of their interfaces. For this reason, we prefer to think of agents as automating diligence, rather than intelligence, for operational challenges.

My site! Rashidazarang.com

Trust yourself to be able to handle agents. Stop trying to be too safe, you’re paying the price with ignorance. Just use Claude Code with Opus 4.5.

Last week, I ran an experiment. Instead of building incrementally, I described exactly what I wanted to ChatGPT 5.1 Pro to draft a prompt, and then asked Claude Opus 4.5 to deliver it in one shot.

The result changed how I think about about speed when working with AI.

Last week, I ran an experiment. Instead of building incrementally, I described exactly what I wanted to ChatGPT 5.1 Pro to draft a prompt, and then asked Claude Opus 4.5 to deliver it in one shot.

The result changed how I think about about speed when working with AI.

People are starting to use AI not just to think, but to feel. When something hurts or feels uncertain, it is easy to offload that discomfort into a system that instantly turns it into clarity: an explanation, a plan, a message, a neatly packaged insight. It works. It feels good. But there is a hidden cost.

If we outsource emotional uncertainty too quickly, we skip the part where we actually feel it. The system digests the discomfort before we do. Over time this can make us excellent at understanding our lives but worse at sitting with the parts of experience that have no immediate answers. We get analysis instead of depth, interpretation instead of emotional endurance.

AI is powerful as a thinking partner, but it becomes risky when it becomes an emotional bypass. Some forms of growth only happen in the silence before clarity. If we replace those moments with instant interpretation, we trade long term resilience for short term relief.

This is not an argument against using AI. It is simply a reminder that some of the most important human capacities develop in the space where no external system can feel on our behalf.

When OpenAI launched Canvas yesterday, many called it a step backwards.

But conversation and spatial interfaces aren’t competing; they’re complementary.

Chat captures intent. Canvas structures complexity.

The real question isn’t which one wins, it’s whether we know when to use each.

Location: Mexico

Remote: Yes

Willing to relocate: Yes

Technologies: Python, TypeScript, React, Supabase, Postgres, Docker, Cloudflare Workers, MCP, WebRTC, agentic AI orchestration

Résumé/CV: https://rashidazarang.com

GitHub: https://github.com/rashidazarang

---

I am Rashid Azarang, a systems architect and builder focused on making intelligence usable. I design and implement cognitive systems that enable human-like interaction through AI agents.

Some recent work:

- Supply Chain Risk Management Platform — turned fragmented data into operational clarity.

- From Sync Bridge to Data Warehouse — re-architected brittle integrations into a coherent warehouse.

- Open Source Twilio SMS Dashboard — practical tooling others have since adopted.

- AWS CloudWatch Interface — lightweight logs explorer with MCP adapter.

I also maintain ChatGPT Exporter (80+ stars) and multiple MCP servers/agents. Previously, I helped scale a COVID-19 testing platform from ~1k to 100k+ tests per month, serving over a million people.

I’m looking for early-stage engineering roles where frontend speed meets backend reliability, especially in real-time systems, or AI.

Agents act. Infrastructure enables. A Navigation Layer enforces constitutional invariants and improves how we improve... Doug Engelbart’s bootstrapping in code.

We’ve had physical, financial, and human capital. All were tools waiting for human hands. What happens when capital itself starts to think?

You know that sinking feeling when a customer says "I never got your text" and you have no idea why? That's exactly why I built Twilio SMS Tracker; an open-source dashboard that shows you everything happening with your SMS messages in real-time.

Open source repo: https://github.com/rashidazarang/twilio-sms-tracker

who ever's downvoting my comment... You're literally trying to shut up the positive review due to your lack of empathy? WTF? This is my preference, I'm an expert in tech and I don't hold your negative views... Stop trying to control the narrative.

I got so excited watching these videos and going through the product page. I completely ignored the price tag without putting any resistance and I thought to myself: I'VE GOT TO HAVE THIS!

Not only that... I started to think about ways I could use this!! I pictured myself using them... I visualized it all, and then remembered when I felt this way when the Ipod was released, and then again, when the first Pebble watch was launched or maybe even, the first kindle.

Although there's going to be some strong competition in the next 1-2 years with Apple, as we all know, the "thin phone" is nothing about the phone, and all about their pathway towards wearables...

I must have this. This is a game changer. WOW!

I’ve been trying to run multiple AI sessions at the same time without everything tripping over itself. This write-up shares the pattern that’s worked: keep each run in its own sandbox (I use Git worktrees), give the small helper agents clear jobs, route messages on purpose, and let CI catch mistakes before they spread.

There's a ton of people that are so resistant to change... Like, finding people who embrace the future and is okay with uncertainty for a while is tough, mostly on scenarios where their age is 35+, though not so dependent on age, but on mindset... I do understand his position, as it's the law of selection... If people are not at least trying to "adapt" to the new reality, then firing them, can also serve as a signal to others... Adapt or die!

Wow! You did something amazing! The first few interactions, I felt as if I was actually spinning a wheel. Great job, bravo!

Yeah, I know LLMs aren’t there yet. But there’s still a lot of room to grow from micro-managing AI by building memory, refining MCPs, and running agents on real work.

People talk about “using AI” like it’s just another app. You sign up, click around, maybe save some time.

But it’s really about building systems that keep learning and improving without me.

I’m going all in on AI so I can eventually go tech-free.

I wrote a short post on something I keep running into:

Not all software is equally easy for AI agents to integrate with.

Some have stable APIs and IDs. Others change identifiers, have no API, or require fragile UI hacks.

MCP is great for standardizing how agents talk to tools, but it can’t magically fix bad or inconsistent software.

If you’re choosing software for agent integration, knowing this difference matters.

I relate, but you need to become healthy. Get up early in the morning, run with your friends or join a group of runners. That’s the answer. Regulate your sleep, get your blood checked, and if you’re low on vitamin D3, supplement it. Follow Bryan Johnson’s blueprint as a reference.

Then go build. Be 50X.

I do relate with what you’re saying, BUT there’s a solution. It’s not a pill, it’s not a diet, it’s not giving up and recruiting people just because you’re lonely.

We’ve spent years building systems that understand... But now, we’re entering a phase where AI can actually act. Not through APIs or complex integrations, but right from your machine, with simple tools and natural language.

In this post, I explore a powerful shift made possible by MCP servers:

https://rashidazarang.com/c/when-ai-gets-its-hands

What happens when the DB gets large? How do you handle idempotency? (What if SQS delivers twice?) The cron job is still a single point of failure...

Yeah, we’ve seen that too. Raw AI output isn’t reliable enough for high-stakes data work.

That’s exactly why we don’t let AI run migrations. We use it to speed up the boring parts, like mapping table structures. But humans are always in control.

Totally agree. We wouldn’t trust AI to run that kind of migration either... And we don’t.

But here’s what we do use AI for: • Mapping legacy schemas • Spotting patterns • Generating boilerplate ETL code fast

Then humans step in: • Validate every mapping • Write custom logic for edge cases • Test everything... every field, every BOM, every relationship • Migrate with deterministic, human-reviewed code

We're using AI to write the boring integration code that moves data from System A to System B. The actual data processing is deterministic code that's tested like any critical system.

Correctness: 100% schema mapping accuracy after human validation. We've never had a data type mismatch or field misalignment make it to production. The AI suggests mappings at ~85% accuracy, humans catch and correct the remaining 15%.

Completeness: Zero data loss incidents. We run reconciliation reports comparing source record counts to destination. Any discrepancy fails deployment. Most common issue: the AI initially missing compound key relationships, which we catch in testing.

Tax/Financial Data: Yes, we handle financial data for several clients, including:

QuickBooks to data warehouse pipelines (invoice/payment data)

Payroll system integrations

Revenue reconciliation between CRM and accounting

Our approach for sensitive data:

AI generates the integration logic, never sees actual records

Test with synthetic data matching production schemas

Run parallel processing for 1-2 cycles to verify accuracy

Maintain full audit logs of all transformations

Human sign-off required before production cutover

We're not replacing deterministic processes with probabilistic ones, that would be insane for production data.

Here's what actually happens:

1. MCP exposes system schemas in a standardized way 2. AI analyzes the schemas and suggests mappings 3. Engineers review and validate every mapping 4. AI generates deterministic integration code (think: writing the SQL, not running it) 5. We test with real data before any production deployment