Pinned Tweet
Introducing the Human Creativity Benchmark.
The first eval that scores AI models the way creative experts do.
406
181
2,141
407,471
Freelancers and agencies who build with Lovable: @contra just handed you free Lovable Pro and a chance to compete for a $25k prize pool.
The brief: create a hypothetical appointment-based business.
1st place: $8,000. 10 total winners. Go build.
4
2
40
3,133
ben retweeted
We recently gave every creative on @contra a US account and routing number, unlocking a simpler way to get paid.
It works like this:
- Create an invoice with bank transfer enabled
- Your client gets an email with your bank details
- They sent a wire from their bank
Within 24 hours, the payment arrives, and the invoice is marked paid. The best part: transfers count toward your Contra earnings and help build your reputation.
1
10
38
1,659
ben retweeted
Guess who made it?
Devin loves building apps on its own PC. Now, it works even better on yours.
Introducing two updates to Devin for Microsoft users: native support for Teams, and a first-party Microsoft 365 integration to connect to your mail, calendar files, chats, and more.
9
1
78
6,253
Thrilled to have @contralabs_ai represented here. @ajnulty and I have some gems to share.
Can't wait to see everyone there.
Reinforcement learning is having a moment.
For the past few months, we’ve been working with @tryterac to put together something special: Convergence — an event for the people building the future of RL.
Speakers from @GoogleDeepMind, @databricks, @mercor, @arena, @joinHandshake, @turingcom, @deeptuneai, @contra, @Stanford, @Princeton, @UCBerkeley, and more.
Our goal: make this THE event for the RL community.
Oct 29 in SF. Come join us → terac.com/events/convergence…
5
3
33
3,344
ben retweeted
Instant 2.0 is here! This is a new way to build for commerce.
If you can describe it, you can build it.
Bundle builder? Check.
Complete Shopify theme? Check.
Product quiz? Check.
High-converting landing page? Check.
We rebuilt the Instant Canvas from the ground up with a code-first architecture designed for AI agents.
The result is a Canvas that's more powerful, dramatically faster, more efficient.
And no and no longer limited to the features we've built for you.
If you can describe it, you can build it.
22
23
184
28,658
ben retweeted
come work with meeeee
@contralabs_ai is hiring research scientists on ai eval and data!
looking for folks who:
> get frustrated to see Fraunces all the time
> know what a GPT restaurant menu looks like and want to fix it
> love to sit down with data for hours
apply in thread
4
22
2,176
ben retweeted
@emonpixels taking full advantage of their Contra bank account details + boosting their reputation at the same time. $250k soon 🔥
This would’ve been $250k long before, but the main issue was not having bank accounts. Most of my payments came from Stripe to my banks (templates alone have generated $100k+).
Most clients were comfortable with Stripe payment links. For clients who didn’t trust me, I used Contra.
Now Contra makes it much simpler. I connected my @Contra US bank account to Stripe, so all transactions go directly to Contra — everything in one platform.
Hopefully, I’ll reach $100k very soon! 🔥
Thanks to @contra @freelancer_matt
1
2
20
1,189
Muse Spark 1.3 will reshape the economics of coding agents.
The quality to cost ratio is exceptional, especially once an agent has a direction and moves from ideation into execution and refinement.
On our Landing Page leaderboard, Muse Spark 1.3 ranked #8 overall with a 42.3% win rate, landing just behind Gemini 3.6 Flash and ahead of DeepSeek V4 Flash.
It is the latest frontier-level model from @AIatMeta.
We also ran it through our Human Creativity Benchmark against Claude Fable 5.1, GPT-5.6 Sol, Kimi K3, and Muse Spark 1.1 across three stages of the creative process:
Ideation → Mockup → Refinement
The results show where Muse Spark 1.3 gets especially interesting: turning an existing creative direction into a high-quality finished product at extremely low cost.
6
12
64
5,742
Despite Muse Spark 1.3 uses similar sections to its predecessor, designers called Muse Spark 1.3 outputs “elegant”, "restrained and professional", and "creative"
while calling Muse Spark 1.1 outputs "old and poor quality," "boring," and "wishy washy".
1
3
242
TLDR: Muse Spark 1.3 is a strong performance-per-dollar model.
Reach for it when the concept is approved and cost matters.
Skip it for Ideation-heavy work, where frontier models are meaningfully ahead. A follow-up study is coming for Meta's Max Effort model soon.
Full card: contralabs.com/research
2
1
5
271
ben retweeted
Yes, direct wires & ACH transfers coming from your Contra bank account details count towards your total platform earnings, boosting your reputation 🚀
This is the easiest way to boost your earnings.
3
7
35
1,011
The Museum Challenge is bringing out real craft 🔥
@AhmadSauleh took a 19th-century Solomon Islands shield from The Met's archive and turned it into a surfboard in @floraai, keeping the pattern, shine, and figure intact.
5
2
24
2,447
ben retweeted
"Hey Astra, make me a Mac app. It should render a 3D iPod that you'll build in Blender, then using the original iPod UI and interactions I want to be able to visualize all my Codex threads on it."
15 minutes later: Boom.
What's even in this model? What time we're living in!??
144
317
6,463
688,601
ben retweeted
if not for things like this, i’d leave tech and be a farmer in rural china
Playgrnd® beta is live.
No account. No paywall. 100% free.
32 tiny tools for making weird, beautiful things, with new ones dropping every day.
Animate in one click. Dither everything. Export SVGs.
Go make something weird. 🛝 playgrnd.tools
1
4
1,321
Since claude is down,
I guess I finally have time to share that Fable 5.1 leads all four dimensions on our Human Creativity Benchmark:
>general preference,
>usability,
>prompt adherence, and
>visual appeal.
Its biggest lead is at Ideation, where models produce a website from the prompt alone.
One designer called the output "the perfect alliance between the prompt and the output," and "client ready in its first look.”
We’re introducing Claude Fable 5.1 and Claude Mythos 5.1.
They're the world’s most advanced models for coding and knowledge work.
8
13
52
7,374
Compared to Fable 5, we also noticed Fable 5.1 producing fewer generic outputs and better first-prompt results.
Hero sections were more diverse and interactive, and the storytelling was stronger.
1
4
158
Fewer generic outputs. Better storytelling. The top spot on our Landing Page leaderboard.
Claude Fable 5.1 is the strongest landing page model we've tested.
Full card here: contralabs.com/research/upda…
4
167