Your favorite tech creator

Technical creator. Involved in everything AI, data, and tech

[Your favorite tech creator @causingheadache]

The edge computing market hit $168B in 2025. IoT devices are headed to 39 billion by 2030.
And yet: most vector database are mot built for air-gapped or latency-critical environments.
Healthcare. Manufacturing. Military. The cloud can't serve them, and nobody's building for them.

[Your favorite tech creator @causingheadache]

@3rdsonx The led

[Your favorite tech creator @causingheadache]

If you were building a coding agent today, which model would you benchmark first?

GLM for complex coding, DeepSeek for cost-sensitive tasks, Qwen for tool orchestration, or Kimi for long coding sessions?

I’d test them on the same repo tasks before choosing.

Chutes just added 7 models across these families to its OpenAI-compatible API.

I like that they included model IDs, use cases, and pricing, so you can compare them with one integration.

https://t.co/aMxbOHZ2rb

[Your favorite tech creator @causingheadache]

@Fanfulladev Agreed, I’m always grateful for open-weights and the control they give us

[Your favorite tech creator @causingheadache]

If you were building a coding agent today, which model would you benchmark first?

GLM for complex coding, DeepSeek for cost-sensitive tasks, Qwen for tool orchestration, or Kimi for long coding sessions?

I’d test them on the same repo tasks before choosing.

Chutes just added 7 models across these families to its OpenAI-compatible API.

I like that they included model IDs, use cases, and pricing, so you can compare them with one integration.

https://t.co/aMxbOHZ2rb

[James @MrJamesKe]

We have normalised things that should make an entire nation furious. Hospitals without medicine, schools without resources, unemployed graduates and politicians living like royalty. This is not normal. It is a national disgrace!

[Meggyyyy @pinkyy_nurse]

If you’ve been following Lauretta’s journey, please keep standing with us. We have just one more chemotherapy session to get through before her break. Your donations, reposts and prayers have carried us this far. Please help us finish this phase together.

[EgoKuwait 🇰🇼 @Egokuwait]

Good morning all, I’ll keep saying thank you because you people have come through in ways I never imagined.
My phone is still buzzing with alerts, I am super grateful🙏
May the universe be kind to you and all your loved ones.

Below is the total amount that have been gotten so far, a total of 5.5m and still counting.
The 1m is not reflecting in my account yet cos it was restricted for some reason, I’m still trying to sort that out.

Also, my Wema bank have been refusing some credit alert.
Please below is my Moniepoint account

Alumona Nancy Chidinma
6578948406
Moniepoint MFB

I have 1.5m more to go.
Thank you very very much
May the good lord in his infinite mercy show up for you.

[Orcabi @orcabi_music]

APC really DAMAGEEEEE this country!

Please on January 16, 2027, vote Peter Obi of NDC✌🏾, for a better Nigeria😭😭🙏🏾

[Your favorite tech creator @causingheadache]

I like how Endform breaks down the quarantine approach here. The test keeps running, failures stay visible, and you can stop it from blocking every PR while you investigate.

https://t.co/VgLjSDBDXg

[Your favorite tech creator @causingheadache]

If your Playwright checkout test fails randomly, but passes every time you rerun it. Do you keep retrying, skip it, or remove it from the merge gate while investigating?

Because at some point, the team starts ignoring the failures entirely.

[Your favorite tech creator @causingheadache]

Came across this Endform guide that walks through the setup with Playwright and GitHub Actions. It also covers where the approach starts breaking down as your test suite grows.

https://t.co/VgLjSDBDXg

[Your favorite tech creator @causingheadache]

A Playwright trick for dealing with flaky tests:

Tag them @quarantine, exclude them from your blocking CI job, and run them separately.

You still get failure reports without every flaky test holding up a merge.

Just don’t forget to assign an owner and a deadline.

[Your favorite tech creator @causingheadache]

Rewire built something along these lines, using Actian VectorAI DB to store reusable failure signatures.

A fix from a brake-pricing error was later retrieved for a healthcare copay error. Great case study:

https://t.co/xG33Zeb5nC

[Your favorite tech creator @causingheadache]

If an AI agent makes a mistake, what should its memory store?

1. The conversation
2. The incorrect answer, or
3. the reason it failed?

For me, I’d store the failure pattern. That way, another agent could recognize the same mistake even in a completely different domain.

[Your favorite tech creator @causingheadache]

If an AI agent makes a mistake, what should its memory store?

1. The conversation
2. The incorrect answer, or
3. the reason it failed?

For me, I’d store the failure pattern. That way, another agent could recognize the same mistake even in a completely different domain.

[Your favorite tech creator @causingheadache]

I like how Rewire implemented this. Their agent proposes workflow changes, validates them through replay, and stores successful patches in Actian VectorAI DB for future reuse.

The architecture is here:

https://t.co/xG33Zeb5nC

[Your favorite tech creator @causingheadache]

Would you let an AI agent rewrite its own workflow?

I would, but only with a regression gate.

Generate candidate fixes, replay historical cases, and promote a patch only if it resolves the failure without breaking existing behavior.

How would you validate it?

[ọsàn wẹ́wẹ́ 🍋‍🟩✨ @tosineiste65214]

Hii hiiiii, people were kidnapped in igando area around lasu.
Some teenagers and an elderly woman, they took turns raping a teenage girl and made the others watch.
They made them hold cardboards and say untrue things about themselves, that they were gay. 1/2

[Your favorite tech creator @causingheadache]

@alexey_anshakov Thanks for this

[Your favorite tech creator @causingheadache]

Would you let an AI agent rewrite its own workflow?

I would, but only with a regression gate.

Generate candidate fixes, replay historical cases, and promote a patch only if it resolves the failure without breaking existing behavior.

How would you validate it?

[Boki @boki_io]

1,401 people saw our StoryChief alternatives article on Google. Zero clicked.

Boki found why: a meta description that was too long and an intro that made readers wait for the answer.

The video shows how we fixed it. We’ll share the results in 4 weeks.

Got an article like this? Reply with the link and we’ll check it 👇