AEKR

Human orchestration engineering by this guy @megalomaniack_.

[AEKR @__aekr]

https://t.co/tOSZuP30uY

[AEKR @__aekr]

@thsottiaux Better QA and application testing / computer user skills. Have better eyes to test flows using clicks, not background VTP.

[AEKR @__aekr]

Your customers aren't on X, here are your competitors.

Enjoy the ride!

[AEKR @__aekr]

Wherever I went, I managed to draw curiosity toward my work.

No matter the role I was assigned, an impulse to feed my curiosity was always a driving factor.

Learning can be addictive sometimes, especially if it opens doors for you.

I'd encourage others to walk beyond their boundaries and encounter inspiration, passion and their own dreams face to face.

You have no idea how many doors are out there waiting for you to open them all.

[AEKR @__aekr]

@claudeai So now we have Teams in here?

I never bought the "AI would kill humanity" bait but this is confirming it.

[AEKR @__aekr]

Your AI agent has a hidden dependency: the person who knows why the code is ugly and works.

Do you have a chunky storeprocedure you want to "enhance" with AI? First check whether the weird exception is a mistake or the reason orders still ship.

Stack Overflow’s 2026 survey released on October 6 got a trustworthy reality check: among respondents answering its context-barriers question, 61% cited important context living in people’s heads and 63% cited incomplete information.

Weird, right? Well, DORA’s 2025 research supported this before: AI amplifies an organization’s existing strengths and weaknesses. AND weaknesses... not magically repair them all.

My two cents: use AI, yes. Absolutely. But capture the decisions, exceptions, and business constraints before asking it to improve things. Then test whether the business still works (not just the code compiles), and make sure your goal isn’t to preserve or enhance "ugly" code when the goal should clearly be to understand what you’re replacing.

The old "if it works as is, don't touch it" is still a reality one must endure. Be careful automating a "new hire" who rewrites everything before asking why it exists, especially if you want to have happy Fridays with no PRs and long weekend hours rolling back Production entirely.

A bigger context window still can’t read your senior engineer’s mind. 😎

https://t.co/c6bFekDeDe

[AEKR @__aekr]

I know this is an old one, but trust me, it's still out there, especially in corporate settings:

AI adoption should not mean giving employees a chatbot and teaching them how to write longer prompts.

If every interaction starts with a "You are a supply chain expert with 20 years of experience..." something is wrong.

Enterprise AI should already understand the domain, project context, engineering rules, previous work, available tools and reusable scripts.

It should manage context, memory, session compaction and orchestration and call specialized agents/tools when needed.

The user should be able to ask: “what’s happening with this container?”… and the framework should know what to do next without looping forever or searching on the web.

A chatbot is an interface, not an AI engineering framework.

That distinction is a big part of what I'm exploring with https://t.co/T7s6Yn5NyR -- moving from prompting AI to orchestrating AI-enabled engineering systems.

Let's see what happens!

[AEKR @__aekr]

Don't be the Jensen Huang of the relationship.

[AEKR @__aekr]

Good morning, America.

[AEKR @__aekr]

Practicing is certainly the greatest laboratory of human thinking.

[AEKR @__aekr]

The next step with AI rests with the people using it: moving beyond subscriptions and experimenting with their own local setups.

Qwen and other open-weight models offer a practical path to get started, with the right hardware and guardrails.

I’d suggest taking a hybrid approach: OpenAI for orchestration, design, QA, and debugging, with a capable local model handling implementation and unit testing.

That’s a combination I’ve been exploring lately and it has been working amazingly.

Encourage y'all safely exploring this with the right setup (never local, always workspace/codespace sandbox).

[AEKR @__aekr]

@MariaGracia40 Just try it for a while and revaluate with time, you might need to adjust the plan a bit, but it will definitely help you to reach your goals.

[AEKR @__aekr]

Regardless what you want or what you do, managing yourself, acknowledging and respecting the plan will get you in the right place.

Leverage can achieve anything, and it begins with You. https://t.co/3wFt1uEpFO

[AEKR @__aekr]

@CarlaAntunez1 I hope so too.

[AEKR @__aekr]

Keep momentum, it will happen.

You're the main source of energy. https://t.co/ZmTFz9z2Mj

[AEKR @__aekr]

@MariaGracia40 Thank you.

[AEKR @__aekr]

@CarlaAntunez1 Thank you.

[AEKR @__aekr]

https://t.co/tOSZuP30uY

[AEKR @__aekr]

@MariaGracia40 Glad you enjoyed it.

[AEKR @__aekr]

No AI will create better than you. Ideas are human realm.

[AEKR @__aekr]

@MorganVonDruitt Very well said.

[AEKR @__aekr]

AI can design your landing page, but you need to leverage judgement to define the right direction and philosophy of your product.

Do not let AI writing all, slop will become uncontrollable.

[AEKR @__aekr]

Tell your boss it is his fault.

Drink that coffee in peace in the break room.

Don't fall for it.

Tell Copilot to read your emails, classify responses per priority and write suggested responses for each one.

Be careful with Copilot in Excel, tho. You better check those formulas yourself, no AI can beat Dany's 30 years old spreadsheet macros.

DO NOT install Claude in your corporate laptop, EVER.

Stop doomscrolling while you are at the office.

Here are some roses for you, and tell HR I said "hi".

[AEKR @__aekr]

Do not consume. Create.

[AEKR @__aekr]

@jack I make better social cards in PowerPoint than this one.

C'mon Jack.