thi4go 🦬 retweeted
🧵 Deli AutoResearch SKILL is now officially open source! 🎉
victorchen96.github.io/auto_…
Alongside it, we’re dropping our 4th survey paper — this time on Self-play.
victorchen96.github.io/auto_…
Inspired by AlphaZero, we got a powerful insight: prior knowledge doesn’t always lift the ceiling.
Models can discover more globally optimal solutions just by playing against themselves.
The biggest change in this paper?
For the first time, the AutoResearch Agent autonomously planned GPU experiments — and submitted actual RL runs on the DeepSeek 285B model.
The entire RL pipeline — experiment design, code writing, running, debugging, and conclusion summarization — was 100% automated, with zero human intervention from me.
This was incredibly difficult, but an incredibly important step.
victorchen96.github.io/blog_…
GRPO is the tool being called by the AutoResearch Agent here.
We see this as the beginning of our Continual Learning research journey. 🚀
As always, this is my personal research project, unaffiliated with any organization. All views are my own.
#AI #ReinforcementLearning #SelfPlay #OpenSource #AutoML #ContinualLearning #DeepSeek
thi4go 🦬 retweeted
The psychological toll on a populace that attempts to adhere to behavioral norms while leaders brazenly, openly engage in corrupt self enrichment is enormous. We are re-writing incentives at lightning pace, eroding the trust which is core to a functional society.
thi4go 🦬 retweeted
Replying to @AnthropicAI
Hi, I'm the creator of a programming language and proof assistant named Bend, with 19k stars on GitHub. We'll launch a V2 soon. Having access to Mythos Preview would greatly help me improve the security of the language, and I believe the language itself will play an important role in software security in the future. Can I get involved on Glasswing?
nitter.cf/VictorTaelin/status/20…
Anthropic claims they won't launch Mythos because it exposes bugs in software, making it too dangerous.
I'm the creator of a new language named Bend (19k stars on GitHub). Its version 2 is coming next month, including a 10x faster CPU and GPU runtime, compilers to 5 different languages, a massive stdlib, and, most importantly, a *complete proof checker*. That makes it the first general language that can prove the correctness of its own programs, so, conveniently enough, it could be the way out of this very mess Anthropic is worried about.
Sadly, Bend2 is now reaching 100k lines of code, making it increasingly hard for us to audit and verify it all. Proof checkers are particularly security-sensitive, because a single bug can lead to false theorems being accepted, undermining the entire trust model of the system. Even Lean, Coq and Agda had bugs in the past. We just finished Bend's initial consistency checker.
Having Myhos audit our implementation would greatly improve Bend's security. In turn, a secure Bend could greatly improve the security of all other software, providing a solution the very problem that prevents Mythos from being released.
I hope this message reaches someone from Anthropic, and they kindly consider letting Bend2 be part of Glasswing!
thi4go 🦬 retweeted
Nano Banana Pro is the biggest unlock for DTC creative since Meta rolled out dynamic ads.
Most brands will scroll past this.
Don't be most brands.
I just generated 50+ campaign-ready images from ONE product photo.
No studio. No photographer. No 2-week turnaround.
(Yes, this image was made with Nano Banana Pro 👇)
This isn't about replacing your creative team.
It's about removing the bottleneck that's been killing them.
What's been strangling DTC creative teams:
✗ Burning $10k on a shoot that fatigues in 72 hours
✗ Waiting 3 weeks for "simple" lifestyle variations
✗ Needing a new shoot for every audience persona
✗ Gibberish AI text that kills your hook
✗ One aspect ratio that doesn't work anywhere else
What Nano Banana Pro actually unlocks:
✓ Upload 14 reference images → 100% product fidelity (no more warped logos)
✓ Generate infinite lifestyle shots from one SKU (virtual photoshoot)
✓ Swap models instantly for persona testing (no talent booking)
✓ Render pixel-perfect text inside the image (legible hooks, any language)
✓ Relight and reskin winning ads to fight fatigue (day to night in seconds)
✓ Outpaint to any aspect ratio without cropping your product
I wrote the playbook.
7 protocols. Real prompts. Actual use cases.
→ The "Anchor Stack" for virtual photoshoots
→ The "Mannequin Swap" for persona testing
→ The "Hook Factory" for native text ads
→ The "Science Ad" for explainer infographics
→ Asset Recycling to fight ad fatigue
→ Global Scale for instant localization
→ Format Fluidity for omnichannel distribution
This isn't prompt engineering. It's creative operations.
Want the full Nano Banana Pro Guide?
👉 Like this post
👉 Re-post
👉 Comment "BANANA"
And I'll send it over.
(Must be following so I can DM)
thi4go 🦬 retweeted
There's probably ten trillion ZEC and no one would know.
Hell, the developers could know about a stealth inflation bug, keep it secret, and cash out over time without anyone ever knowing.
The public would have to independently discover the inflation bug themselves, which is unlikely since there are a handful of people in the entire world who understand zero knowledge proofs and they're all under the umbrella of some corporation with signed NDAs.
It's an impressive money printing scheme but a scam nonetheless.
thi4go 🦬 retweeted
Several people have asked for an overview of $DCR StakeShuffle decentralized p2p mixing used by Bison Wallet:
1) Parties that want to mix coins create ephemeral identities to use in mixing session and submit a mixing pair request to participate in a mix to the p2p network
1/7
thi4go 🦬 retweeted
Decred represents the best of crypto $DCR
thi4go 🦬 retweeted
Nature is Conscious and is Aware of Everything
Trees become motionless at the sound of the flute. They perceive the Sound Frequency—they are listening.
✨🙌🏾💫
© manifiesto_do

























