Blixt @blixt

Building @biscuit_so • Previously @Framer, @Spotify biscuit.so AMS/BCN Joined August 2007

Tweets

7K
Followers

4K
Following

2K
Likes

28K

Biscuit @Biscuit_so

12 hours ago

A recipe app that pulls live nutrition data. A content dashboard with latest YouTube view counts. That kind of data comes from external APIs. Connecting one on Biscuit is just a conversation. 🧵 Here's how it works - example of YouTube Data API to show view counts on a video:

1 1 3 155 1

View Details

Blixt @blixt

a day ago

@gdb I flew too close to the sun

0 0 4 317 0

View Details

Blixt @blixt

2 days ago

@venturetwins @mo_norouzi @ideogram_ai Fun and fast model, really enjoying good text adherence. Added to Biscuit!

1 0 2 316 0

View Details

Blixt @blixt

2 days ago

Creating with Biscuit means you have something real from the very beginning, the rest is up to you!

Biscuit @Biscuit_so

2 days ago

Every app you build on Biscuit comes with its data out of the box. 🎉 No database setup. No empty placeholder screens. Your screens fill with realistic content the moment you build. Here's how 🧵

1 3 18 2K 10

0 0 5 565 0

View Details

Blixt @blixt

2 days ago

@jurrehoutkamp @AntonDrukarov @framer The Framer logo has 10 sides, 5+5 is 10, GPT-5.5 confirmed.

0 0 3 103 0

View Details

Blixt @blixt

2 days ago

@MengTo @benhylak But it's still janky tbh, many times it does (correctly) need to jump out to my main Chrome because of browser-use limitations or cookies, but I haven't found any way to avoid an approval popup every few hours, and Chrome steals focus etc. Still early days!

0 0 0 11 0

View Details

Blixt @blixt

2 days ago

@MengTo @benhylak I have this set up under Personalization and it's completely solved the confusion for me.

1 0 2 82 1

View Details

Blixt @blixt

2 days ago

“only create functions if shortening a 300+ LoC function or deduplicating logic” ☝🏼 GPT-5.5 *will* add 5 layers of functions instead of modifying an existing function, so without telling it the above your code will become a mess (unmanageable by GPT-5.5)

0 0 3 168 1

View Details

Blixt @blixt

2 days ago

Two instructions that work for me to avoid GPT-5.5's overabstraction and tendency to keep duped code: “remove any dormant code” ☝🏼 it seems to respond well to the word ‘dormant’ and successfully removes unused code etc

1 0 6 357 5

View Details

Blixt @blixt

2 days ago

Also pretty wild that 15 minutes later I can have a decent video draft of something fun

Blixt @blixt

2 days ago

Pretty impressive that gpt-image-2 can one-shot this

0 0 3 499 0

0 0 7 305 0

View Details

Blixt @blixt

2 days ago

Pretty impressive that gpt-image-2 can one-shot this

0 0 3 499 0

View Details

Blixt @blixt

3 days ago

@marias_martin @ishverduzco @Biscuit_so A true superhero!

1 0 0 17 0

View Details

Blixt @blixt

3 days ago

Agreed! 5.5 has such a different vibe from e.g. 5.3-codex

keith @keithwhor

3 days ago

gpt-5.5 writes bad code imo. too many abstractions. tons of small helper functions for no reason. fairly lazy. it is fast and prototypes working stuff well. but code falls into disarray quickly. opus-4.8 is like the opposite. super slow. overthinks. but pristine code.

1 0 5 838 0

0 0 3 304 0

View Details

Ives van Hoorne @CompuIves

3 days ago

It took blood sweat and tears to build this, but every time I see the screens streaming in, I realize it was worth building it 🥹

11 5 182 22K 123

View Details

Biscuit @Biscuit_so

3 days ago

Launching your app and hoping for the best is not a strategy. You need to track, analyse, and improve - based on real data. In Biscuit, you just ask. 💬 The AI already knows your users, your content, your app. The insights are richer because the context is already there. 📊 Try now 👇

2 2 7 802 0

View Details

Blixt @blixt

3 days ago

I first learned about kanban in 2009, but back then I actually had to do the work myself after moving the task! We really are living in the future.

Ives van Hoorne @CompuIves

4 days ago

Making good progress on tasks! You have a central "cofounder" chat that creates tasks for you, then every task a ticket (plan + comments) that gets implemented & tested by sub-agents. Next up: night shift!