@nedo

making cool tech for boring businesses to do more than humanly possible.

CTO @

[@nedo @nedo_sy]

Every now and then I meet someone who is amazing at selling visionary talk.

They are also borderline morons and unwilling to learn by themselves.

Yet the market listens to them and puts them on a pedestal.

And every business probably needs one.

[BridgeMind @bridgemindai]

FABLE 5 CAME BACK NERFED.

We re-ran the July 1st version of Claude Fable 5 on BridgeBench.

The results are brutal:

Debugging: 86.2 → 25.9
Refactoring: 73.6 → 38.4
Hallucination: 75.9 → 61.7

The new guardrails are kicking in on way too many tasks and falling back to Opus 4.8.

This is not the model that got banned.

Anthropic owes everyone an explanation.

[@nedo @nedo_sy]

It's 2026 and everyone is in the AI race and AirAisa still can't build a decent chat bot...
Or maybe I just hate Bo

[@nedo @nedo_sy]

Every morning I write code. No agents. Just meat fingers hitting the keyboard and manually saving files the old fashioned way !

manually crafting structure, code layouts, boundaries, logs...

Then I run @mattpocockuk /grill-with-docs skill.

and I'm enjoying every bit of it !

[@nedo @nedo_sy]

Google The Gell-Mann Amnesia Effect.

Every tech enabled business right now should make it mandatory for every founder / exec to memorize this term by heart.

[Adam @adamdotdev]

It's 2026 and your CEO just sent you a 2,400 line pull request.

You get a cup of coffee and sit down to review it.

It's a disaster. A dozen unrelated refactors. Unused methods with names like `convertFromBase10` and `normalizeBeforeSerialization`.

You catch a few hardcoded API keys, but that's ok. It's part of the dance. They didn't consider that someone might look at this diff. Here's a comment buddy.

They respond in an hour (after Copilot, qodo, CodeRabbit and Greptile finish their reviews) saying we shouldn't worry about "implementation details" anymore, those are relics of the past. Hey let's jump into a room and figure it out. We can't just agree to disagree, this is probably my last job in tech and I can't watch this fucker burn the place to the ground.

The PR merges and goes to prod. You feel a shared sense of apathy and dread with Hannah the intern (she has to review his AI generated social media posts ever since Grok got too imaginative).

That night you go to sleep and have nightmares of that code. You can still see the shapes of it on the backs of your eyelids.

You go to work the next day ready to quit. You no longer understand the system. There is no foundation. Time to use those savings and an SBA loan to buy a liquor store and never login to GitHub again.

[@nedo @nedo_sy]

Cognitive debt > technical debt

[@nedo @nedo_sy]

PJ drivers don't follow the same road rules / etiquette as the rest of KL...

[@nedo @nedo_sy]

Personal experience so far:

Codex > Claude Code
GH copilot > Claude Code
GPT 5.5 > Opus 4.7

American tech bros are loyal to Claude Code but also critical of Anthropic ...

It's so weird, I don't get it

[@nedo @nedo_sy]

The more I use AI and the deeper I investigate what other people in tech are doing with it... The more I realize that AI is not just a tool, it's a succubus that seduces humans into a productivity trap and digs us deeper into the dunning kruger effect.

[@nedo @nedo_sy]

The jump from building something specific to building something generic isn’t incremental, it’s exponential.

I built plenty of ETLs and data pipelines so building a generic engine with solid conventions felt like the natural next step.

Doing it right is unbelievably difficult.

[@nedo @nedo_sy]

I'm happy that my productivity is on an upward momentum with all the of the AI assisted coding.

But I'm feeling absolutely knackered and definitely burning out much faster with the amount of code that I have validate, review, restructure because of it...

[@nedo @nedo_sy]

I had a data integrity issue this morning with some records...
So I started writing a SQL patch to fix it.

the query grew too complex to ensure it corrected the data.

I asked codex gpt5.4 high with plan mode to complete and proof it.

Now I have 2 data integrity issues 👌🏻

[@nedo @nedo_sy]

It's almost comical how many LLM wrapper startups are competing on price and offering "credits" that obfuscate tokens.

Your entire business will melt down like an ice cube when the era of subsidised tokens comes to an end.

I'm waiting. Patiently for that "in hindsight" moment.

[@nedo @nedo_sy]

If every hot new startup is using LLMs to generate strategy, write code, do outreach, lead gen, hiring, etc..

How the hell do you differentiate your business from your competitors? Assuming that your "edge" is generated by a 20$ LLM subscription 🙂

[@nedo @nedo_sy]

Free LLMs + doom scrolling are the perfect storm for cognitive decline on massive scale...

I see it almost everywhere now...

[@nedo @nedo_sy]

"Move fast and break nothing" in software is a marketing ploy, it's not real

[@nedo @nedo_sy]

I'm also spending on Cursor and GH Copilot.

5.3-codex high is peak in my experience.
I find code generated from this model is easier to fix during review compared to code generated via Opus.

My observation so far is:
Codex is thorough and slow. Opus is noisy and slow-ish

[@nedo @nedo_sy]

20$ sub on both Claude-Code and ChatGPT:

GPT-5.3-codex > Opus 4.6

That's by a huge margin !

My experience using skills, project references, MCPs, etc...

Codex is far more superior on TS / Node projects. Opus misses the mark almost every time...