Post

More from @hasantoxr

Hasan Toor

@hasantoxr

Feb 27

Holy shit... Someone just built the ultimate prompt collection for AI image creators.

It's called MeiGen, it scrapes the hottest prompt posts from X every week and curates them in one place.

No more bookmarking 50 tweets. No more losing that prompt you saw 3 days ago.

100% free. 100% Open Source.

Here's the problem MeiGen solves:

The best AI image prompts live on X.

But they're buried in your bookmarks, your likes, your "I'll come back to this" pile that you never come back to.

MeiGen pulls the hottest ones weekly curated, organized, searchable.

What you get:

→ Weekly curated prompts trending on X (real engagement, not random)
→ Filter by model like NanoBanana Pro, GPT Image, Midjourney
→ One-click generate + save to your collection
→ Real view/like counts so you know what actually performs

The top prompt right now? 159K views on X.

Read 6 tweets

Hasan Toor

@hasantoxr

Feb 24

Holy shit... Someone finally benchmarked AI code reviewers on REAL pull requests.

Not demos. Not vibes. Actual F1 scores across 8 tools.

The gap between #1 and #8?

34 percentage points.

Your eng team is probably using the wrong one right now: ↓

@entelligence benchmarked 8 tools on real pull requests.

Every "which code review tool should we use" convo ends the same way:

→ Someone shares a demo
→ Someone shares a vibe
→ Nobody has real numbers
→ Team picks based on brand recognition

That era is over.

The full F1 score breakdown across 8 tools:

🥇 Entelligence — 47.2%
🥈 Codex — 45.4%
🥉 Claude — 42.8%
4. Bugbot — 39.4%
5. Greptile — 36.9%
6. CodeRabbit — 33.0%
7. Copilot — 22.6%
8. Graphite — 13.4%

Copilot at 22.6%. Graphite at 13.4%.

These tools are being sold to your eng team RIGHT NOW.

Read 7 tweets

Hasan Toor

@hasantoxr

Feb 24

I just ran 6 AI tools in one workflow.

No API keys. No dashboard switching. No glue code.

Just typed what I wanted and Claudcode + SkillBoss executed the entire thing.

Here's the breakdown: 👇

Most AI setups look like this:

→ ChatGPT tab for writing
→ Midjourney for images
→ Runway for video
→ Zapier to connect them
→ 4 API keys you'll inevitably break
→ 2 hours wasted before you even start

That's not a workflow. That's a second job.

SkillBoss changes this completely.

100+ AI models & skills in one workspace.

Chat. Image. Video. Audio. Email. Scraper. PPT. Storage.

Claudcode orchestrates all of it you just give the instruction.

One prompt. Full execution.

Read 7 tweets

Hasan Toor

@hasantoxr

Feb 21

RIP to every dev team charging $50K to build an internal dashboard.

UI Bakery just made every internal tool your dev team ever built look like a waste of time.

It's called UI Bakery, it builds and deploys a fully functional internal app in 2 minutes.

No sprint. No Jira ticket. No engineer bottleneck.

Here's how: ↓

Here's what it actually does:

→ Connect to 45+ databases (Postgres, MySQL, MongoDB, Snowflake, Redis, OpenAI...)
→ Describe the app you want in plain language
→ AI Agent generates and deploys a fully functional app
→ 2 minutes. Production-ready. SOC 2 compliant.

Not a prototype. A real app on live data.

The features that actually matter:

- 80+ pre-built React components use anything, no restrictions
- One-click deploy with auto-scaling, SSL, CDN included
- Built-in RBAC, audit logs, MFA enterprise security out of the box
- Self-host option for air-gapped environments
- React code export zero vendor lock-in. Ever.

Read 7 tweets

Hasan Toor

@hasantoxr

Feb 19

🚨BREAKING: The open-source Alexa killer just dropped.

It's called OpenHome, a smart speaker dev kit you can run AI agents on.

No Amazon. No Google. No vendor lock-in. Your data stays local.

100% Opensource.

OpenHome @OpenHome runs on Raspberry Pi 4 (8GB RAM) with:

→ Far-field AEC mic array (hears you across the room)
→ Full-range speaker with deep bass
→ Local LLMs + local STT/TTS (fully offline capable)
→ Smart home control via Matter/Thread & MQTT

Your agent finally has ears, a voice, and a home.

Every AI agent today is stuck in a chat window.
OpenHome gives it a physical body.

Vibe code voice abilities in 10 minutes with the Ability Editor.

Clone your voice in 60 seconds.

OpenClaw agents already run on it out of the box.

This is the missing layer for AI agents.

Read 5 tweets

Hasan Toor

@hasantoxr

Feb 17

BREAKING: Someone quietly built the first AI assistant that runs on a $5 chip.

It’s called MimiClaw and it's a full @OpenClaw style agent running on an ESP32 microcontroller without @Linux

It’s built with local-first memory and privacy by default.

No Linux. No Mac mini. No Raspberry Pi. No VPS.

100% Opensource. MIT License.

The setup is wild:

→ ESP32-S3 board ($5-8)
→ No Linux
→ 16MB flash, 8MB PSRAM
→ 0.5W power draw (runs 24/7 off USB)
→ Talks to you through Telegram
→ Uses Claude's API with full tool use (ReAct agent loop)

That's a full AI assistant for less than a coffee.

But the memory system is what makes this insane.

It stores everything as plain text files on the chip's flash:

- SOUL .md → the bot's personality
- USER .md → your preferences
- MEMORY .md → long-term memory across reboots
- Daily notes + full chat history

Your data never leaves the device.

Read 6 tweets

Share this page!

Enter URL or ID to Unroll

Hasan Toor

Try unrolling a thread yourself!

More from @hasantoxr

Hasan Toor

Hasan Toor

Hasan Toor

Hasan Toor

Hasan Toor

Hasan Toor

Did Thread Reader help you today?

Don't want to be a Premium member but still want to support us?

Send Email!