Will Norris

A biological intelligence

[Will Norris @WillNorris2]

@frankdotlee Thanks a lot! I did spot that, but assumed it was a work around rather than default workflow. I’ll give it a go! Appreciate you getting back to me :)

[Will Norris @WillNorris2]

Hey @frankdotlee! I'm testing out Amplitude MCP. Loving it so far. Qq - are there tools for editing charts, or is this deliberately restricted?

[Will Norris @WillNorris2]

@jasonfried @avinashjoshi @dhh +1 to Hey CLI!

[Will Norris @WillNorris2]

@1littlecoder @_juliushere @fal Thank you very much!!

[Will Norris @WillNorris2]

@1littlecoder @_juliushere @fal OK thanks for the update. Would be amazing to have a proper status page where can monitor this more transparently. Do you have an ETA for the full fix?

[Will Norris @WillNorris2]

@1littlecoder @_juliushere @fal NB seems to have been purged. Just Flux not working still.

[Will Norris @WillNorris2]

@1littlecoder @_juliushere @fal fal-ai/flux-1/schnell, fal-ai/nano-banana & fal-ai/nano-banana/edit all seem to have 5k+ length queues right now. Anything that can be done there?

[Will Norris @WillNorris2]

@1littlecoder @_juliushere @fal Hey @1littlecoder small follow up here. Any update?

[Will Norris @WillNorris2]

@1littlecoder @_juliushere @fal Seems this is the issue... Very long queues needing to flush. Do you have a status page or something I can monitor?

[Will Norris @WillNorris2]

@fal Hi! Do you plan to add support for 512px images?

[Will Norris @WillNorris2]

@cutler_max @JeremySternLA @patrick_oshag Lovely

[Will Norris @WillNorris2]

@hehe6z Congrats on the launch! Would love to see this on @fal

[Will Norris @WillNorris2]

@IanMcclanan Well done mate! Looks fantastic

[Shreyas Gite @shreyasgite]

Famously, founders think that copying Elon means going super ambitious and big. Which is true for PR but not engineering.
Elon is quite scrappy with technologies. For early iterations he takes tech that's proven commercially and ready for deployment, not something at the frontier.

As Kyle Vogt from Cruise (now bot company) put it nicely: you want to bet on deploying technology that's commercially proven, not something at the frontier. Because you never know how long it's going to take to make that commercially useful - 1 year or 10 years.

Let's take self-driving. Tesla's early efforts in 2016 started with absolute focus on imitation learning along with lots of supervised learning - tons of data annotation for bbox, segmentation etc. In 2016, the hype was around unsupervised learning and RL. So doing supervised/imitation learning was a very boring approach.

The robotics startup (bootstrapped, <10M funding) betting on world models as main breakthrough might not be the right idea. Solving 1-3 boring problems via some derivative of pick and place wouldn't be that bad of an idea.
Like Vivek from Matic Robots puts it: solve something simple and add a buy button to it.

[Will Norris @WillNorris2]

Hey robot, grab me the kind bar

[Will Norris @WillNorris2]

I always feel like a dumbass when I ask a reasoning model what I think is a complex question and it replies after 5 seconds of "thinking"

[Will Norris @WillNorris2]

@matbogus In my experience it’s the model which applies diffs between current and generated code which doesn’t work with notebooks.

It’s surprising because I would have expected the cursor research team to use notebooks as part of their dev workflow…

[Will Norris @WillNorris2]

@IanMcclanan Smart robots!

[Will Norris @WillNorris2]

First time seeing explicit memory in cursor... Love it!