Marcel Losso

Desenvolvo umas coisas legais, mas provavelmente inuteis.
Ah, e tambem trabalho como Software Engineer mas isso é só um detalhe

[Marcel Losso @marcellosso]

was reading brown's rust book and loved that you can highlight and leave notes while reading. wanted that on my own site so I built penciled

one line, annotate('#article'), and readers can highlight + take notes. inspired a lot by how clean and simple @emilkowalski's sonner is

doesn't touch your dom at all, uses the css highlight api

[Marcel Losso @marcellosso]

Watching it fail in public: https://t.co/N1jlH3R5nr

[Marcel Losso @marcellosso]

I put a monkey on my GitHub profile. Every day it smashes 10M random keys and looks for Hamlet.

[Marcel Losso @marcellosso]

Open an issue with a banana emoji → +1M keystrokes now, monkey dumps what it typed, your name goes on the wall.

[Marcel Losso @marcellosso]

I think most indie devs don't have a marketing problem. They have a 'shipped once and went quiet' problem.

I know because I do it. Every product I've launched got exactly one post.

[Marcel Losso @marcellosso]

@jackfriks just tell it to buy more ram, easy fix

[Marcel Losso @marcellosso]

@arvidkahl For a one-liner I still do it by hand, anything bigger than that > agents. I think the agent pays off on checking the side effects

[jQueryScript @jqueryscript]

penciled.js: a 15 kB open-source JavaScript library for persistent text highlights and notes on any web page.

Built on the CSS Custom Highlight API.

👉 https://t.co/t7KxrVBupD

[Marcel Losso @marcellosso]

@jqueryscript Thanks for the feature! Happy to hear any feedback 🙏

[Marcel Losso @marcellosso]

~15kb, or 6kb headless if you want your own UI. react adapter included

playground: https://t.co/zbwHFcudz0
repo: https://t.co/a9lHYceAy4

[Marcel Losso @marcellosso]

was reading brown's rust book and loved that you can highlight and leave notes while reading. wanted that on my own site so I built penciled

one line, annotate('#article'), and readers can highlight + take notes. inspired a lot by how clean and simple @emilkowalski's sonner is

doesn't touch your dom at all, uses the css highlight api

[Marcel Losso @marcellosso]

@SurfNewsBrasil Eu tambem tinha 🤡🤡 Bom que tirei print do meu Rank 1 la, pq jaja vai sumir KKKKK

[Surf News Brasil 🌊 @SurfNewsBrasil]

Atualização do Fantasy! 5 pessoas até o momento. Ninguém quer concorrer aos prêmios durante as etapas, a lycra do Filipe Toledo e ajudar o ADM???

Link para a assinatura https://t.co/Q541dsUUYK

[Marcel Losso @marcellosso]

@markoilico Pleaseee!!

[Marcel Losso @marcellosso]

@TheMattBerman @ridgewallet ADS

[TemporaMap | See Any Place in Any Year @temporamap]

Saint Petersburg, 1905 - Colorized.

[Kling AI @Kling_ai]

Kling Omni Launch Week Day 1: Introducing Kling O1 — Brand-New Creative Engine for Endless Possibilities!
Input anything. Understand everything. Generate any vision.

With true multimodal understanding, Kling O1 unifies your input across texts, images, and videos — making creation faster, smarter, and more effortless.

Limited-Time offer available for subscribers. And stay tuned for more surprises!

For the next 12 hours ONLY
Follow, Like & Retweet to get 200 Credits — for everyone who participates!
PLUS, 200 lucky winners will snag a *1-Month Standard Plan*
All sent straight to your DM

[Marcel Losso @marcellosso]

@marclou I think it would be cool to be able to connect with datafast, so potential buyers can see the analytics

[Marcel Losso @marcellosso]

Essa imagem foi criada no temporamap e eu to impressionado, a unica coisa que eu consigo ver que entrega que eh IA seria o "hilux" ali

[Marcel Losso @marcellosso]

Still lots to optimize, but today was a fun learning moment, and the results are finally close to where I want them.

https://t.co/uxZf2DsEUW

If you want to try it out, I’d love your feedback.

[Marcel Losso @marcellosso]

A good prompt makes all the difference.

I’ve been tuning TemporaMap, and with the right prompt engineering… this happened.

Sharing some findings 👇

[Marcel Losso @marcellosso]

The real catch:
Sending this huge JSON directly to NanoBanana made it ignore important instructions deeper in the structure: lens, DoF, angle, everything started looking a bit… rubbery like the image below.

The fix?
I still generate the JSON, but now I programmatically convert it back into clean natural language before sending it to Nano. Instant improvement. All the “after”s images above were generated after this change