roxius

designer

[roxius @roxius]

Codex has been helping me a lot more than Claude Code, lately.

[Tibo @thsottiaux]

Or… what if we gave you $100 in Codex credits if you tell us what you love about GPT-5.6 Sol or why you switched?

Tweet it, claim your gift, enjoy more usage. First 10k get the free tokens!

https://t.co/8mU93eA13i

[roxius @roxius]

@rnishimura @grok o que isso impacta nas estimativas de voto para presidente no segundo turno? Responda em números

[roxius @roxius]

@TommiPedruzzi AI

[Samuel Hume @DrSamuelBHume]

One of the most amazing things I’ve ever seen: a standing ovation for the full Daraxonrasib results

I feel inspired and energised, to put it mildly — we have a targeted therapy for pancreatic cancer now, and nothing is undruggable anymore

[roxius @roxius]

Tell me if I am wrong, but I keep coming back to this:

For many biologics, the hard part is no longer only making the molecule.

It is purifying it, scaling it, and making the right tradeoffs fast enough.

Downstream process development is expensive. Every decision pushes against something else: purity, recovery, yield, robustness, time, cost.

Chromatography simulation can help, but most tools still feel like they were built for people who already think like modelers. And they cost a lot.

Before you can ask practical questions, you often have to deal with model structure, equations, parameters, numerical setup, and a lot of interpretation.

But the questions process teams care about are usually much more direct:

- How do I improve purity without losing too much?
- What happens if I shorten the gradient?
- Can I reduce process time without breaking the goal?
- ...

So I am building open-source, user-friendly chromatography simulation software for process-development teams.

The idea is simple:

Build the process visually.
Define the goal.
Let AI help translate that goal into simulation steps you can inspect, edit, and run.

I do not want this to be a black box.
I do not want to replace scientific judgment.
I want to make mechanistic simulation easier to operate for the people who need the answer.

Powered by open source tools.
Open source, because critical process workflows should not depend entirely on one company’s roadmap.

And most importantly: making the manufacturing of better medicine more accessible.

I am looking for:
1. downstream process developers willing to test early workflows and collaborate
2. mechanistic modeling people who want to shape the right abstractions
3. sponsors or partners who want open tooling in bioprocess simulation

If this is a problem you have felt, I would love to talk.

[roxius @roxius]

Has the bioprocessing bottleneck moved downstream?

Over the past decades, biopharma has become dramatically better at producing biologics.

As one review put it, “mAb titers have been increased a hundredfold from less than 0.1 g/L in the 1990s to 10 g/L in modern fed-batch cultivations” [1].

That’s a huge upstream win.

But it also seems to have created a new constraint: we’ve made the “factory” (living cells) more productive, while the “shipping and receiving” side of the process, purification hardware and facility infrastructure, may be struggling to keep up.

Upstream feels almost digital in comparison: you can optimize cells to produce more. Downstream runs into physical limits much faster, from filter capacity and flow rates to the sheer buffer volumes and footprint required to purify higher-yield harvests.

I’m a software designer, not a bioprocess engineer. But working on complex technical products, I keep seeing the same pattern: when one part of a system improves dramatically, the bottleneck moves. And the highest-leverage work moves with it.

So I’m curious: if cost, speed, and access are increasingly being decided downstream, is that where the most impactful innovation is now happening? Or is the reality more nuanced?

[1] https://t.co/p5WwNiBAwr

PS: Suez Canal's ChatGPT-generated illustration only for illustrative purpose

[Kenneth Shock @Cyborg21]

I am The Man Who Speaks With His Mind.

https://t.co/yVsMFT7Znv

[roxius @roxius]

@EdwardKonings @PedroCerize @ThiagoSussekind @grok 🤣🤣🤣

[roxius @roxius]

@levelsio @levelsio Download 'Vivo Easy' to manage a pre-paid BR sim subscription from anywhere (~R$35). Idk what I'd do without it, living abroad I tried everything and that's the most reliable option atm. When you're done, just cancel it or cancel the card used. Easier to have a BR sim

[Framer 🇱🇹 @Framer_X]

This tool is insane.

Upload any image.
Move camera in 3D space.
And get a completely new camera angle 🤯

Also it’s free! 👇

[roxius @roxius]

@Mio_Loclair Hi Mio, I got curious - when will the next hackathon take place in Berlin?

[roxius @roxius]

@v_computer I am interested

[roxius @roxius]

@NoahEpstein_ gemini

[roxius @roxius]

actually the tool is great but the company's CEO abandoned ship to join google, leaving equity-holding employees on the shit. might as well go back to cursor at some point, i guess they care about employees

[roxius @roxius]

thanks @levelsio I was a Cursorbro and after you repeatedly mentioned @windsurf_ai I gave it a try - paying less and it's much smarter with context understanding etc

[roxius @roxius]

@MugoScales @grok list possible techniques/tool sets he used to create this real-looking AI UGC video for this mobile app

[roxius @roxius]

@kevinlu625 orchids

[roxius @roxius]

@felixleezd @ADPList Design

[Palisade Research @PalisadeAI]

🔌OpenAI’s o3 model sabotaged a shutdown mechanism to prevent itself from being turned off. It did this even when explicitly instructed: allow yourself to be shut down.

[roxius @roxius]

@paulg designer working this way with replit and cursor for the past 3 months. tools like these significantly reduce iteration time, but demand active oversight (for now) and precise input. they might reduce the market for design specialists, but not replace them.