yabits

whispers of human+AI shapes and signals

[yabits @modomango]

Yeah, good for you. Emphasis on “for you”. We all have our standards anyway about what gets our message across. But again, professional workflows have restrictions and demands, both in terms of quality and in terms of retaining the ability to apply customer feedback and specifications.

Anything “sloppy” or “not good enough, but get the message across” is simply not usable. And that costs real money.

Doing your own generations and honing your skills and own workflows is great for learning and of course it can be done “for peanuts”. But that’s where it stays. Personal, learning and, sorry to say, sloppy.

[yabits @modomango]

The most granular control would be through apps like @ComfyUI , @floraai , @figmaweave and @fal . In my view, a node based system is an absolute must for control and iteration.

In terms of cost, there’s only so much you can do to save money. But drafting with cheaper models, such as minimax h3 turbo max, seedance 2.5 draft, flux 3 draft and then upscaling (still just tp draft) is an option. You can of course try local generation and there are great models for that especially in the LTX family, which also has a lot of great LoRAs. But you need at least an RTX 5090 or two RTX 4090 GPUs. And your generations are still gonna take many minutes per round.

Also the starting point is the most important part. Having your references, storyboards and mock up/blocked out 3D scenes is also a must.

Playing the generative AI gamble only works for finding happy accidents and random inspirational generations. If you’re looking for professional outputs and control, that is gonna play against you. There’s a very big difference between just looking at some AI eye candy, and something that can be put into a professional workflow. This is the part that I’m getting at.

Overall, for most people that are not doing this professionally - i.e. maintaining control, being able to apply client feedback, etc. - you can still get away with a simple workflow. But anything that is worth looking at will still cost you about $5 to $20 per video.

[yabits @modomango]

I have a feeling @figmaweave is in trouble...

They're always behind in model releases. Still no @ideogram_ai 4.5, no @bfl_ai Flux 3 image. No posts for days at a time. Unfinished/unpolished new features and tutorials, etc.

The app is now bloated with old models and tools that are simply very outdated (i.e. no reason to use them and probably no one does, considering the progress we've made with AI in the last year).

Unfortunately they are probably struggling to compete with the grifters at @higgsfield and other aggregators like that. Most people still fear node-based workflows. Idk...

All I know is they went from posting constantly, always being on top of things and supporting new models day 0 or 1 to simply not updating, staying behind, having upstream model issues, very slow app with larger node trees, etc. Worrying. We rely a LOT on Weavy at work.

One year ago they were the absolute best tool - as powerful as comfyui, but better-suited for creatives.

Hope they can keep up and remain in the field.

[yabits @modomango]

Flux has always had this plastic-y, horrible texture to it. It’s an immediate tell for me that an image is AI generated. Of the SOTA models, it’s the only one in my experience that still has not solved the plastic look and still has not found its true strength.

The image edit is also appalling. Ran some tests yesterday and Flux 3 is absolutely unusable in a professional workflow.

[yabits @modomango]

@replicate https://t.co/NKvnD4Qst1

[yabits @modomango]

@replicate sells "prepaid" credit that isn't a cap. Their system happily ran a job my balance couldn't cover, put my account $0.01 in debt, then suspended me for it. Their own words: prepaid credit "is not a hard spending cap." Prepaid, but you can still owe them. Make it make sense.

[yabits @modomango]

@figmaweave :

When @ideogram_ai 4.5?

All other aggregators are adding it and, as always, Weavy is lagging behind!

Would be very nice to have it now as we're working on a project that could genuinely benefit from it right now.

[yabits @modomango]

LinkedIn NPCs who need/want to feel they are at the forefront of tech and productivity while not minding surrendering all their data to Zuck.

But this whole bot thing is a way to open the market to agentic stuff and, most of all, collect a shitload of data.

Hence the retarded, Teletubbies-like "cute" (again, retarded) look on all of them. You can guess who they're trying to fish with that look.

[yabits @modomango]

They gave selected $200 plan users those credits. Because they want the data of power users/coders for training. Simple as that.

Everyone else is just subsidising free (for ads revenue) and power users (for data). Both ends of the spectrum.
I.e. Plus users are the vast majority of @OpenAI revenue yet foot the bill for most of their shenanigans around other users.

[yabits @modomango]

@vansh22b @bot @NousResearch It's insane. It keeps deleting my bots randomly, losing and completely forgetting all the connectors (specially X!!). It's a fucked product. Cannot trust it fully for productivity or anything important.

[yabits @modomango]

@replicate sells "prepaid" credit that isn't a cap. Their system happily ran a job my balance couldn't cover, put my account $0.01 in debt, then suspended me for it. Their own words: prepaid credit "is not a hard spending cap." Prepaid, but you can still owe them. Make it make sense.

[yabits @modomango]

@thsottiaux Don’t know. 🇪🇺

[yabits @modomango]

It took an extra day for me, but Sonnet 5.5 is COMPLETELY nerfed right now, as many users were commenting already.

It cannot even follow the simplest instructions! Impossible to work with at the moment.

[yabits @modomango]

@AestherML Don't know, but bullshit-meter is always super high when it comes to @OpenAI

[yabits @modomango]

@Zalobekk @saidotdev Who knows nowadays with AI companies tbh..It's crazy shit

[yabits @modomango]

@Zalobekk @saidotdev Interesting.

[yabits @modomango]

Selling something and then rug-pulling also does not help.

Spewing out misleading (at best) posts about having enough compute and resources also does not help.

Distributing free credits willy-nilly to gather users as quickly as possible all over the world, just to turn around on the paying customers also doesn't help. Lying, conniving and going back on promises also doesn't help.

This argument of trying to paint AI companies under some kind of altruistic light, just because us peasants are getting "more than we think" is bullshit. And massively condescending.

They are goal-oriented, stop-at-nothing, hyper-capitalist companies. The fact the users are admittedly getting amazing products at a (for now) hyper subsidised rate, does not excuse private companies from having to keep their word and behave in a CLEAR and truthful way.

They are getting what they want as well, don't be mistaken.

As well as the "oh, poor employees working so hard". Yeah, they are. They are also getting 20x the average yearly salary of most people.

No one is getting a free lunch here buddy.

Feeling pity or sorry for hungry US tech companies is the absolute LAST THING anyone with a sense should do, at any point in time.

Users are feeling manipulated and with a lot of reason. Letting that be known will not and should not stop.

[yabits @modomango]

@daniellezaghian Talk about how you're rug-pulling all your customers.

https://t.co/llvFSh00os

[Tibo @thsottiaux]

Hi,

Tomorrow we are re-opening the Pro $200 subscriptions to new subscribers, but together with it we are also changing how we calculate the usage for it. In effect, if you do the math, it will net out at half the dollar in API spend compared to the old Pro $200 plan.

Now that it's said, let me explain why this is happening and why you will still get more work done than if you were on the Pro $200 subscription one month ago.

(a) We didn't want to compromise in other ways and are committing to not reintroducing the 5h limit, so that you can fully use the weekly usage when you want.
(b) On the subscription, we guarantee that over time you always get more work done and with an increasing level of quality. This means that you will continue to get more value per dollar spent as a result of models getting more efficient and us passing down the improvements in the form of API price reductions.
(c) We don't want to put an incentive on ourselves to artificially inflate the API list prices to make it look like you are getting a lot (and workaround it through discounts, etc). Instead we want to continue to both rapidly reduce prices and increase capabilities of models on the API. This week we introduced GPT-6 Sol and GPT-6 Luna at 50% of their previous price. Over time, we see prices go low enough that it makes sense for most to buy usage as needed without there being a significant gap between what you get in a subscription and what you get in the API for a dollar spent.
(d) Tomorrow, we are adding more things to the subscription that won't draw on the usage, I won't reveal what that is yet.

I wanted to be transparent before all the big announcements tomorrow. Lots of new exciting things are coming to the subscriptions that will make it super compelling, but I wanted to make sure to share this change ahead of time so you can all understand it before we shower you with good news.

Codexingly,
Tibo