Shanda Group, founded in 1999 by Tianqiao Chen, was an early internet gaming pioneer in China. Today, it invests globally in deep tech, AI and frontier science.
Thank you Laure Belot (@curieusedetout) at @LeMondeFR
for this thoughtful article about the Chen's approach to philanthropy. What an interesting series to be featured in. https://t.co/s3rpGduKR4
How are you using MiroThinker the most?
1️⃣ Vote in the poll below
2️⃣ Share your detailed workflow in the comments
💰 GIVEAWAY: The top 5 most-liked comments sharing use cases or feedback by April 8, 12:00 PM will EACH receive 2,000 MiroThinker Credits!
The final day of Memory Genesis Competition 2026 is April 4th.
Compute gets the headlines. Memory is where intelligence actually lives.
Keynotes. Roundtables with OpenAI + AWS. $80K+ in prizes. Live demos on stage.
📍 Computer History Museum · Mountain View Spots are limited. Register now. ↓
Final chapter of Memory Genesis 2026 is here.
April 4 · Computer History Museum · Mountain View
Finalist teams demo live. 30+ judges score in real time. $80K+ in prizes.
Keynote on agents, reasoning & memory.
3 expert panels. Awards ceremony.
The future of AI memory gets built on stage.
https://t.co/LgZnb1ETud
Proud to see @miromind_ai 's MiroThinker paper ranked #1 Paper of the Day on @huggingface Face — a meaningful recognition for a team that has been quietly doing serious work.
As @tianqiao_chen notes, the milestone is not the point. The direction is. Building AI that doesn't just sound right, but actually is right — that's a harder problem, and one worth solving.
Congratulations to the entire MiroMind team. This is just the beginning.
Getting ranked #1 Paper of the Day in huggingface is nice.
But that’s not the point.
The point is this:
Most AI today is optimized for conversation.
Very little of it is built to solve real problems.
And the gap is not just scale.
It’s verification.
Without verification, AI doesn’t produce answers.
It produces guesses that sound convincing.
That’s the problem we’ve been working on with MiroThinker.
•long-horizon reasoning
•structured planning
•and verification at both local and global levels
Not a chatbot.
But a system designed to actually work — in science, finance, and other high-stakes domains.
Still early. Still imperfect.
But this is a direction we believe matters.
Curious how others are thinking about verification in AI systems.
https://t.co/aQncyQj2nZ
#AI #AgenticAI #Reasoning #AIResearch #MiroThinker
Another strong result from the Shanda research ecosystem — this time from @evermind.
MSA (Memory Sparse Attention) scaling to 100M tokens is a meaningful technical step. But as @tianqiao_chen points out, the bigger idea is the shift from context as input to memory as a system — a fundamentally different way of thinking about how AI retains and reasons over information.
Proud of what the EverMind team is building. Keep going.
MSA (Memory Sparse Attention)
→ scaling to 100M tokens
→ addressing the long-standing trade-offs between scalability, precision, and efficiency
What this points to is not just longer context,
but a shift from “context as input” to memory as a system.
This is still early, but if this direction holds,
it could fundamentally change how we think about intelligence in AI.
This work comes from EverMind — our sister team to MiroMind, focused on long-term memory.
https://t.co/32QuJqcK3q
#AI #Memory #LongContext #Architecture #EverMind #MiroMind
Some weeks remind me why building research teams is worth it.
This week, our Tokyo AI Research Institute received great news:
multiple papers from the team have been accepted to CVPR 2026.
Huge congratulations to the researchers and students involved:
Xuangeng Chu, Ruicong Liu — for their work on joint generation of listener and speaker facial expressions.
Haiyang Liu, Yiyi Cai — for their research on natural-language-guided motion generation.
Erwin Wu, Xuangeng Chu, Haiyang Liu — together with collaborators from Tokyo Institute of Science, for their work on dialogue-driven speech–motion generation.
Zhixiang Wang, Zhenghui Huang — for their work on hybrid rendering for world models.
And special recognition to Dr. Kaipeng Zhang, whose team contributed four papers in the world model area accepted to CVPR.
Together these contributions span digital humans, motion generation, and world models — the exact research directions we set for the Tokyo institute from day one.
The Tokyo AI Research Institute operates alongside MiroMind as part of the broader research ecosystem under Shanda.
We chose Tokyo intentionally.
Japan is the global center of animation, virtual characters, and digital entertainment culture.
Our ambition is to build the AI technologies that will power the next generation of games, digital characters, and interactive worlds.
Seeing the Tokyo team’s work recognized by CVPR is an exciting milestone — but it’s only the beginning.
If you are a researcher or engineer in Japan working on digital humans, world models, graphics, generative AI, or interactive systems, we would love to connect.
Great things are just getting started.
https://t.co/s49Pl05Ra4
#CVPR #DigitalHumans #WorldModels #GenerativeAI #AIResearch #TokyoAI
@miromind_ai is building the team to match the mission.
Scientists from xAI, Meta FAIR, Princeton, Caltech, NTU — all focused on one thing: AI that isn't just plausible, but provably right.
This is why Shanda backed MiroMind. Proud of what this team is becoming. 🧠
Big news: three world-class scientists join MiroMind's core research team.
🔬 Dr. @SimonShaoleiDu (ex-xAI, FAIR, Princeton) → Reasoning Models & Training
🤖 Prof. Bo An (NTU) → Runtime & Agent Systems
✅ Dr. @KaiyuYang4 (ex-Meta FAIR, Caltech) → Verifiable AI Lab
Our goal isn't to build a more eloquent AI. It's to build one that's provably right.
📎 [https://t.co/0PyMSReeKt]
Big milestone from our portfolio company @miromind_ai.
MiroThinker-1.7 and H1 are out — built for the problems that actually matter: long-horizon reasoning in science, finance, and engineering. Not small talk.
This is what Shanda invests in. 🔗
@LidongBing Proud moment for the Shanda ecosystem. MiroMind represents exactly the kind of patient, mission-driven AI development we believe in — building systems that reason, not just respond.
Congrats to @LidongBing and the entire team on the launch. 🎉
🧠 Context windows are temporary. Memory is permanent.
The next frontier of AI isn't just generation—it's retention, reasoning, and evolution over time. That's why Shanda is backing @evermind's Memory Genesis Competition 2026.
Alongside @OpenAI and top AI organizations, we're calling on builders to create the memory infrastructure that will define agentic AI.
Tracks: Agent+Memory, Platform Plugins, OS Infrastructure. Prizes: $80K+, revenue share, fast-track opportunities.
This is where the future gets built. Join us.
Jan–Apr 2026 | Global & Hybrid Register now below:
True value lies in solving real-world friction—Aging, Energy, Climate—not just simulating conversation. 🌍
Our founder @tianqiao_chen defines the critical shift from "Liberal Arts LLMs" to "Science LLMs."
This move from Simulation to Discovery drives our entire investment thesis. 👇
AGI isn’t about sounding right. It’s about being right in the real world.
I frame this as “Liberal Arts LLM” (simulation) vs “Science LLM” (discovery)—where causality + verification loops matter most.
🎥 https://t.co/pJebyVmsqY
📄 Text version: https://t.co/xk61xIOGQr
Read the latest writing from Shanda Founder, Tianqiao Chen. His series examines AI in unexpected ways. "Genesis of Silicon: The Flood, Noah, and the Entropy of Tolerance" https://t.co/c8PqErE4wB
We're happy to support @CordanceM and their noninvasive focused ultrasound platform that safely and temporarily opens the blood-brain barrier to deliver therapies for brain tumors, etc https://t.co/dYmO2gR4pO
Congratulations to Li Ye, PhD (Stanford/HHMI) — winner of the 2025 SfN Tianqiao & Chrissy Chen Young Investigator Award! 🎉
His tools (CLARITY, CAPTURE, HYBRID, CATCH) are transforming how we see the brain, linking activity, genes & behavior. 🧠✨ https://t.co/rgUfitUTqY
Award supported by @ChenInstitute.
#Neuroscience #Innovation
🎬✨ The Chen Institute Grand Mirror Studio is redefining science communication with AI-generated art + storytelling. From Beijing to Oscar-qualifying festivals, their work is reaching millions worldwide. Check it out on @Youtube https://t.co/c1Wh8MwB42