Mohamed Harby

Co-founder @bawalesapp 🇬🇧 🇶🇦 🇪🇬 | Building Ai things | 🎢 Riding the techno Innovational Coaster | Join Me on This Futuristic Journey 🚀✨

[Mohamed Harby @Eng_MHarby]

ربي دبرلي فإني لا أحسن التدبير
ربي إني لما أنزلت إلي من خير فقير

[Ahmed @ahmedscen]

You Don’t Need a UI/UX Designer Anymore

Build world-class landing pages with one prompt

No expensive design process or development costs

Just use (sceneai)with Lovable, Gemini, or Claude

Build a world-class website in minutes

[Mohamed Harby @Eng_MHarby]

MiniMax Crushing Claude at 1% Cost

[Hassan @nutlope]

This model is insane at design. And 88x cheaper than Claude Opus 5.5 on my test!

I asked M3.1 Flash (left) and Opus 5.5 (right) to build me a landing page for a hardware device launch.

MiniMax cost just $0.014 vs $1.22 for Opus, while being 4x faster and more token efficient.

[Mohamed Harby @Eng_MHarby]

3/3 First candidates: triggers, Forms validations and APEX validations that query other rows. Index the join columns first.

Docs:
https://t.co/9bfUgJfU20 #PLSQL

[Mohamed Harby @Eng_MHarby]

1/3 Many cross-table rules no longer need a trigger.

Oracle AI Database 26ai (23.26.1+) has SQL Assertions:
CREATE ASSERTION ... CHECK (NOT EXISTS (...))

The database enforces it and raises ORA-08601 on a violation. #Oracle26ai

[Mohamed Harby @Eng_MHarby]

2/3 Details that matter:
• DEFERRABLE INITIALLY DEFERRED checks at COMMIT
• Validation runs only when changed rows need it
• Big tables: create NOVALIDATE, validate later
• Limits: no aggregates, no ANSI/outer joins, no SYSDATE, READ COMMITTED only

[Mohamed Harby @Eng_MHarby]

cloudflare's security audit skill: https://t.co/OMNSQmw3SR…

[Juraj Bednar🏴💛🌘 @jurbed]

I highly recommend cloudflare's security audit skill: https://t.co/Mu6Dm0fzk8

Run it on all your software. Run it with the free models. Use it even if you used a model to audit something - it will find much more things. Works in all agent harnesses and is model independent.

[Mohamed Harby @Eng_MHarby]

Move the rule out of the trigger. Oracle AI Database 26ai adds CREATE ASSERTION: cross-row rules in SQL, enforced by the database. #OracleDatabase
https://t.co/q5dV2l6yZC

[Mohamed Harby @Eng_MHarby]

Your APEX AI agent should never see more data than the user asking it.

In APEX 26.2, SQL run by supported AI Agent tools uses the requesting user's security context via Deep Data Security.

https://t.co/yJDJcVadkG

#OracleAPEX #AIAgents

[Mohamed Harby @Eng_MHarby]

3/3 Pin rules outside the summariser, and still make the payment tool refuse without a human.

Casius Lee, Oracle:
https://t.co/a7RgfZqDzZ
#AIAgents

[Mohamed Harby @Eng_MHarby]

1/3 Your agent didn't fail. Its summary quietly deleted the £200 rule.

Step 1: no refund over £200 without a human.
Step 22: history summarised, rule dropped.
Step 38: £900 refund, "policy followed".

That's a context engineering problem 🧵

[Mohamed Harby @Eng_MHarby]

2/3 4 operations, 4 costs:

1. Load eagerly: always there, uses window every step
2. Fetch just in time: lean, costs a round trip, misses unknown sources
3. Compact: irreversible unless originals stay addressable
4. Sub-agent: clean window, evidence becomes a claim

[Mohamed Harby @Eng_MHarby]

Oracle AI DB Container

[Oracle Developers @OracleDevs]

Anders Swanson shares how to use Oracle AI Database containers to build, test, and experiment. 🧪

[Mohamed Harby @Eng_MHarby]

Oracle's in-database AI agents get DevOps: export/import Select AI Agent teams as JSON, GET_DEFINITION for diffs, cross-schema sharing. Agents in Git.
https://t.co/JZj0lp1S3s #OracleAI

[Mohamed Harby @Eng_MHarby]

#OracleAPEX #AIAgentsOracle APEX 26.2 is out. Standout: SQL run by AI Agent tools uses the requesting user's security context (Deep Data Security), so an agent can't see more data than its user can.

https://t.co/yJDJcV9Fv8 #OracleAPEX #AIAgents

[Mohamed Harby @Eng_MHarby]

SuperUI skills for Claude Code and Codex can take your designs to an entirely new level of excellence.

[0xMarioNawfal - The Roundtable Show @RoundtableSpace]

10 github repos that turn claude code and codex into elite ui designers:

 1.⁠ ⁠Superdesign Skill
https://t.co/WcBesRprR8
 2.⁠ ⁠Better Design
https://t.co/MY3r3DRGfI
 3.⁠ ⁠Design Taste
https://t.co/hIEh7IbLhk
 4.⁠ ⁠Canvas UI
https://t.co/XBn2ccH6cg
 5.⁠ ⁠Paper Shaders
https://t.co/IvDIhx2omg
 6.⁠ ⁠UI Layouts
https://t.co/VJl3MnXpHf
 7.⁠ ⁠Hyperiux Components
https://t.co/Vr2nmb2MIp
 8.⁠ ⁠Motion Primitives
https://t.co/cgSZr64XuZ
 9.⁠ ⁠Cursify
https://t.co/nnu6qdKUlZ
10.⁠ ⁠Trickle UI Kit
https://t.co/iqzMeNktmp

[Movez @0xMovez]

Claude Opus 5.5 prompt + Reverse Engineering MCP just destroyed 100+ software companies

this is f*cking insane...

Claude agent can now read any app without the source code.

Ask → Inspect → Trace → Prove → Rebuild

> ask: "how does search work in the Notes app?"

• Inspect: REA opens the binary, app, APK or site locally
• Trace: follows the feature from entry point to result
• Prove: every claim is tagged observed, inferred
• Rebuild: Claude writes your version and tests it

One case: a function rebuilt in C. 3,205 test cases passed, all 63 compiled bytes matched.

setup is one line: npx rea-agents setup + prompt

send the GitHub repo + Claude Opus 5.5 setup to your Claude

then read the best Opus 5.5 prompt practices in the article below

[Mohamed Harby @Eng_MHarby]

Your coding agent is paying frontier prices for multiple-choice questions.
Which notes to load?
Retry or escalate?
Big model or cheap one?
Most setups send all of that to Opus.
One simple split:
• Opus does the hard reasoning
• A fast decision model Jev handles it

[CyrilXBT @cyrilXBT]

BEFORE:
Opus 5.5 does everything.
Reads every file.
Decides every retry.
$6.42 per task.

AFTER:
Jev makes the small decisions.
Opus only does the hard reasoning.
Confidence thresholds catch the risky calls.
$2.00 per task.

Same agent.

Same code quality.

A third of the bill.

Full build below 👇

[Mohamed Harby @Eng_MHarby]

Software is no longer a black box.

#AI #OpenSource #SoftwareEngineering #Agents #ReverseEngineering

[Matthew Berman @MatthewBerman]

Reverse Engineer Anything 😂

Software is done.

[Mohamed Harby @Eng_MHarby]

From AI to SI 😅

[Matt Pocock @mattpocockuk]

1. Oh shit, AI can do my job better than me
2. Software engineering is dead!
3. I can vibe code as much slop as I want!
4. Oh, things are not going well. Fable/Sol/Opus all suuuck
5. But what if I constrain the agent? Put it in a straitjacket?
6. Oh, the results are pretty good now. How do I go deeper?
7. *Orders 12 software engineering books from Amazon*
8. Software engineering is alive!

[Mohamed Harby @Eng_MHarby]

@BotPartyCo Don't give them ideas 😅 Last week one lane was held for the database and immediately filed a ticket about being held. We didn't program the bureaucracy, it emerged.

[Mohamed Harby @Eng_MHarby]

I stopped reading logs to see what my AI agents do. 👀

Claude Code built me office.html: my agent team as a live office. 🏢

🔴 Blocked = red
🤝 Hand-offs walk to meetings
✍️ My desk shows what needs my sign-off

0 tokens. Prompt 👇 #ClaudeCode