Tweets

Here’s the app on @wabi.

There are fields that let you change sections of the prompt - e.g. age, ethnicity, outfit, location - and upload a product photo. 

wabi.ai/@justine/custo…

Ovis-Image is out on Hugging Face

a 7B text-to-image model specifically optimized for high-quality text rendering

These three devs built Neurana because they were sick of the drag between idea and release: APIs to connect, auth to configure, infra to wrangle, yet nothing shipped.

Here’s how they killed that drag and what you can steal for your builds.  🧵👇

.@SpenserSkates has spent more than a decade building Amplitude from a YC startup into a public company, and in that time, he's had to reinvent himself just as much as the product.

Joining the @LightconePod, he talks through the shift from founder to large-company CEO, the skepticism his team initially had toward AI, and the moment they realized the next wave of analytics would require a full reset.

He walks through the hard reorgs, the bottom-up experiments, and the mindset changes that let Amplitude move fast again.

03:40 - Embracing AI at Amplitude
11:00 - Product roadmap, AI native priorities
15:32 - Org changes & hierarchy
18:40 - “AI killing SaaS”, changing the Amplitude roadmap
23:30 - The “features, not companies” debate, advantages for incumbents
32:14 - Finding good mentors, hyper focus
38:09 - The difference between being a founder vs a big company executive

POV: You're building an agent and it keeps giving weird answers because your PDF parsing is broken 🫠

This is a great walkthrough by @mesudarshan showing exactly how to use LlamaParse to fix this—from basic setup through advanced configs.
The video walks through:

· Why most PDF parsers fail on complex layouts (tables, charts, multi-column text)
· Using the LlamaCloud playground to experiment
· Real demo: parsing "Attention Is All You Need" paper with different settings
· Cost-effective vs agentic vs agentic plus modes—when to use each
· Preset configs for invoices, scientific papers, forms (they tune the parsing prompts for you)
· Advanced options: OCR, language selection, choosing your LLM (Sonnet, GPT-4, etc.)
· Saving custom configs so you don't have to re-tune for similar docsyoutu.be/mUHPPBbumIs?si…w

Check out @tuanacelik's new tutorial showing you how to equip Claude with a proper semantic understanding of Excel sheets!

Claude is a great model for many things, and I know some lovely people who work at Anthropic.

But I just can’t comprehend the constant posturing around AI safety. 

If you truly believe you’re building such a dangerous weapon that everyone should be scared of…why keep doing it?

🚀@SpaceiumInc just launched its first orbital refueling demo on SpaceX's Transporter-15.

Mission teams have been saying the same thing: fuel constraints are forcing real compromises in capability and mission design. Operators know where they want to go, but fuel limits cap maneuver profiles and reduce reach.

Spaceium's precision robotic actuator, now in orbit aboard D-Orbit's ION, is the first step toward solving this. Built in 5 months, it's engineered for real operational environments and designed to evolve into a complete fuel logistics architecture.
spaceium.com/spaceiums-orbi…p

Another test of Kling O1: turning an average car into Lightning McQueen

Paper: arxiv.org/abs/2511.20597