NotebookLM: different approach to AI research.

Upload your sources - PDFs, docs, notes. AI answers based only on what you gave it.

Minimal hallucinations. If it's not in your sources, it won't make things up 🙌

Bonus: podcast-style discussions about your docs. Weird but useful.

Not everything I want to try is available yet.

On my list:
→ v0 by Vercel
→ Codex by OpenAI
→ Sora 2 (waiting for Poland launch 😅)

What's on your "to try" list?

Some ideas are hard to explain. Easier to show.

Lovable: concept to prototype fast. Great for validation.

Worry: non-technical folks trusting it too much. But as a Wix killer? Could be huge 🙌

Curious where this goes.

Eleven Labs started as a curiosity. Now it's part of my workflow.

Transcripts with speaker separation. Custom speech generation.

Fun fact: founded by two Polish high school friends frustrated with bad movie dubbing 😄

A year ago this was sci-fi. Now it's Tuesday.

I used to think voice-to-text was gimmicky.

Then I tried Wispr Flow.

Now I use it for posts, journaling, messages, even prompting LLMs. Talking is faster and more natural 🙌

Small tool, big shift.

(referral: https://wisprflow.ai/r?RAFAŁ34)

English isn't my first language. Writing docs and worrying about grammar at the same time slowed me down.

Grammarly changed that. Typos, punctuation, tone - it handles it. I just write 🙌

Any non-native speakers relate?

Spent 3h turning meeting notes into action items

Today: uploaded to Gemini → "extract action items" → 5 min, clean list

3h→5min. Not everything is code. Know your tools!

What task are you doing manually?

Every company has scattered knowledge - docs, wikis, Slack threads.

Lucky my company uses Glean - I connect it to Cursor via MCP. Coding with company context 🙌

Catch: it's enterprise-only.

Any accessible alternatives for smaller teams?

Code reviews shouldn't be a bottleneck. But they often are.

Graphite changes that. Stacked PRs, faster reviews, cleaner history.

Recently acquired by the Cursor team 👀

Git became the standard. Graphite might be next.

Claude Code handles the work I used to postpone.

Debugging, code reviews, docs - things that break my focus. Now it runs in parallel while I do something else.

Second brain that doesn't get tired of the boring stuff 🤯

What do you delegate to AI?

Cursor Pro is my daily driver for coding.

Switch between Claude, GPT-4, whatever fits. MCP integration connects it to external tools and data.

Not replacing thinking - removing friction between idea and implementation.

What's your daily driver?

A year ago, engineers wrote code line by line.

Today? Cursor, Claude Code, Codex - everyone's experimenting.

"Vibe coding" is no longer niche. I'm still figuring out the right balance.

Where are you on this spectrum?

Just read through ai-2027.com

Fascinating predictions, but feels too optimistic. Timeline seems aggressive.

Then again - a year ago I didn't expect to code with AI agents daily. So who knows 🤷

What do you think?

12 years writing code. Thought I had it figured out...

AI just reshuffled the deck. New game, new rules 🤯
I'm excited about it! 🤩 Going to share what I learn here - tools, experiments, what actually works.

Come along if you're navigating this too! 🚀

There is a critical vulnerability in Cursor!
Fix your IDE before it's too late 🚨

oasis.security/blog/cursor-security-flaw