DEV Community: Neil Okikiolu

Stop Writing Your Own Validators

Neil Okikiolu — Wed, 10 Sep 2025 18:30:00 +0000

Every AI dev eventually hits this wall:

json.decoder.JSONDecodeError: Expecting ',' delimiter

So you write a validator. Then another. Then you realize half your project is glue code just to catch malformed outputs.

1. The Validator Tax

Here's what "rolling your own" usually looks like:

A dozen try/except json.loads() blocks
Manual type checks (if not isinstance(data["age"], int))
Retry loops with time.sleep
Ad-hoc logging sprinkled everywhere

Now multiply that across every agent, every project. That's the validator tax.

2. What If It Were One Line?

Instead of reinventing validation every time, what if you could just do this:

from agent_validator import validate, Schema

schema = Schema({"name": str, "age": int})
result = validate(agent_output, schema, retries=2)

That's it.

3. Features That Replace Your Boilerplate

Agent Validator handles the ugly stuff for you:

🔍 Schema Validation → define with plain Python dicts
🔄 Automatic Retries → exponential backoff & timeouts
🔧 Type Coercion → "42" → 42, "true" → True
📝 Logging → JSONL logs with automatic redaction
☁️ Dashboard → optional cloud monitoring for teams

4. CLI Included

Not just a library — it ships with a CLI:

# Validate inputs
agent-validator test schema.json input.json --mode COERCE

# View recent logs
agent-validator logs -n 20

5. For Indie Devs & Small Teams

Free tier → SDK + local logs
$9/mo Starter → indie devs get cloud logging & dashboard
$29/mo Pro → small teams with alerts & templates
$99/mo Team → larger projects with RBAC & analytics

6. TL;DR

Stop burning hours on custom validators.
Drop in one SDK, validate everything, move on.

👉 Install today:

pip install agent-validator

Repo: github.com/agent-validator/agent-validator

Why JSON Mode Fails (and How to Fix It)

Neil Okikiolu — Tue, 09 Sep 2025 18:30:00 +0000

LLMs are powerful… until you ask them to return JSON.

Suddenly, you're staring at this:

{ "name": "John", "age": 30, "email": "john@example.com"

Missing bracket. Invalid JSON. Pipeline crashes.

Sound familiar? Let's break down why this happens and how you can actually fix it.

1. Why "JSON Mode" Isn't Enough

Some LLMs (like OpenAI) let you request response_format="json". Sounds perfect, right?

Except:

Models still hallucinate comments or stray text.
Outputs truncate mid-string when token limits hit.
Complex schemas (nested dicts, lists) still break.

"Guaranteed JSON" isn't guaranteed.

2. The Hidden Costs

Most devs end up writing custom validators:

Regex hacks to strip trailing commas
Manual try/except loops with json.loads
Silent None returns that mask real errors

You spend hours debugging glue code instead of shipping features.

3. The Real Fix: Schema Validation + Retries

Instead of hoping the model behaves, enforce a schema and automatically retry malformed outputs.

Here's how with Agent Validator:

from agent_validator import validate, Schema, ValidationMode, ValidationError

schema = Schema({"name": str, "age": int, "email": str})

try:
    result = validate(
        agent_output,
        schema,
        retries=2,
        mode=ValidationMode.COERCE
    )
    print("✅", result)
except ValidationError as e:
    print("❌ Validation failed:", e)

STRICT mode → exact type/shape match
COERCE mode → safe conversions ("42" → 42, "true" → True)
Automatic retries with exponential backoff
Local logs for debugging

4. Observability Included

Every attempt is logged to ~/.agent_validator/logs/ with correlation IDs.
No more "why did this break?" at 2 AM.

Need monitoring? Turn on cloud logging:

export AGENT_VALIDATOR_LOG_TO_CLOUD=1

and you get a dashboard of validation attempts.

5. TL;DR

JSON mode ≠ reliable JSON
Stop hacking validators by hand
Use a schema + retries → your pipeline won't break on malformed outputs

👉 Try it:

pip install agent-validator

Repo: github.com/agent-validator/agent-validator

Investors Just Wasted $13B on Anthropic Because They Don’t Get AI Is Regional

Neil Okikiolu — Wed, 03 Sep 2025 12:30:00 +0000

Every AI pitch deck right now uses the same line:

"We're building the AWS of intelligence. Foundational infra. Picks and shovels for the AI gold rush."

That framing is why valuations are sky-high, OpenAI, Anthropic, Mistral, xAI, everyone's being priced like they'll own the entire world's cognition layer.

But here's the reality: AI isn't SaaS. It's strategic infrastructure.

🌍 No Country Will Outsource Its Brain

Would China let an American telco own its 5G towers?
Would Russia let a U.S. company run its power grid?
Would India let foreign firms control its satellites?

HELL NO. Those are national sovereignty issues.

AI is the same. A nation's "thought layer", how its population, military, and corporations interface with cognition, is too important to outsource.

That's why:

China is banned from OpenAI and they are building Baidu/Alibaba/Huawei stacks.
The EU is funding LLaMA distillations and Mistral.
The UAE launched Falcon.
India is openly talking about AI sovereignty.

🪓 The Coming Fracture

So let's be blunt: the "global monopoly" story is fantasy.

Best case: U.S. firms dominate the U.S./allied bloc (NATO, Japan, South Korea).
Real case: the world fragments into blocs, each with their own sovereign stacks.

Think telecom, energy grids, or even chip fabs. Strategic, bloc-bound, heavily regulated.

🤪 Why Are Valuations Still Insane?

Because capital markets don't care, yet.

They want exposure to AI at any price.
Scarcity of investable names means money piles into the same 5 - 6 firms.
Investors are betting they can IPO or flip before the physics hit.

It's the dot-com playbook: sell the dream now, let the next guy 🫵 eat the hangover.

🧨 The Knife Edge

Once sovereign distillation stacks mature, cheap, local, trained on domestic data, the "picks and shovels" narrative dies.

OpenAI doesn't become the AWS of cognition. It becomes the American AWS of cognition. Still huge, but not global.

And that's a very different valuation story.

🔑 Takeaway

If you're a dev or founder, don't build assuming one company will run the world's AI layer.
Build for a fractured landscape: U.S. bloc, China bloc, EU bloc, Middle East bloc, Global South bloc.

The winners won't just be who trained the biggest model. They'll be who own distribution, compliance, and trust inside their bloc.

😊 Hot-take over. Curious what you think: Do you buy the dream of a single AI brain for humanity, or the reality of many smaller, walled-off brains?

Links

Twitter: https://x.com/neil_okikiolu 👈 follow me

Stop leaking .env files - try Secretsnap (free CLI)

Neil Okikiolu — Sun, 31 Aug 2025 07:51:22 +0000

If you’ve ever:

emailed .env files
pasted secrets into Slack
or accidentally committed them to GitHub...

…then you’ll get why I built this.

secretsnap is a lightweight CLI:

secretsnap init
secretsnap bundle .env
secretsnap unbundle secrets.envsnap

That’s it, your secrets are safe.

Free: local encrypt/decrypt/run.
Pro: push/pull bundles, team sharing, audit logs.

👉 Install:

curl -sSL https://get.secretsnap.sh | bash

Landing: https://secretsnap.dev

GitHub: https://github.com/secret-snap/cli

Feedback welcome, especially if you’ve had .env pain in CI/CD.