🚀 Feedback & Showcase: Share Your Most Hardcore MiroThinker Use Cases!

[johnnyise]

Hi all!

If you want to tackle complex problems or make predictions about key events, try out the latest version of MiroThinker 1.5, released by MiroMind AI.

While it may not feel as instantly responsive as some GPTs, every step involves deep thinking and bold verification! Let’s push this model to its limits! Show off your use cases in the comments below.

🔍 Haven't tried it yet? Click the link below to get started!
👉 https://dr.miromind.ai/

📸 Showcase Your Cases!
We want to see how you are pushing the limits of this model. Please share your research results below and show the community how you are unleashing MiroThinker's potential.

Suggested Sharing Format:

• Task: What did you ask it to do?

• Result Display: Please share a Link to the result, a Screenshot of the Reasoning Trace, or a short screen recording of the final output.

• The "Wow" Factor: What was the most impressive part? (e.g., a specific source it found, a logical pivot it made, or the depth of the final report).

We also welcome any feedback on your experience. Feel free to leave your comments below and start the show! 👇

[zoroisnice]

问了下：获得2024年度TGA最佳动作游戏，其第三回中在打颜⾊为白色的飞龙的场景，所取自现实建筑原型与另外哪几个建筑并称中国现存三大辽代寺院？感觉一次任务可以多步思考和推理，搜索整合几百条信息，给出准确答案。

[johnnyise]

真的好玩儿，用来预测1月底的比特币价格！期待中！

帮我预测比特币在2026年1月底的价格区间，最好控制在200美金之内。
https://dr.miromind.ai/share/96d9dad6-ec36-4291-adb2-d0686cd7dde9

video.mov

[hellen-wong]

mirothinker的预测隔天应验了，闪迪大涨20%。

prompt：从连板的角度分析，选5只美股明天涨10%的股票

[hellen-wong]

X上发现一个好玩的

[hellen-wong]

至少加强了我等韭菜的信息获取能力

[btwo163]

2. 如果你是「中性/平衡型」投资者

可以直接采用前面给出的「标准分档」模型：

建仓：

1. 130–135：小仓建仓 20%

2. 120–125：标准加仓 30%（合计 50%）

3. 110–115：若真给到，敢再加 20–30%（合计 70–80%）

4. 跌破 110：不再加，除非你对阿里的长期逻辑极度有信心

减仓：

• 150–155：减 20%

• 160–170：再减 30%

• 180–200：再减 50%，或清仓

这套策略的优点：

• 下有支撑：主力成本会集中在 120–130 附近，只要公司没出根本性问题，出大亏的概率相对较低。

• 上有弹性：如果最终市场接受 180–200 的估值，你有充足空间逐步锁定利润。

3. 如果你偏短线/喜欢波段

那建议你主要围绕 130–155 做波段：

• 130–135 买，150–155 卖
  不太纠结长期 AI 云的故事，更利用情绪和技术波动。

• 必须注意：阿里现在是「消息股+情绪股」叠加

  • 财报、政策、AI 新动作、闪购补贴消息都会带来 5–10% 的日内波动。

• 止损要更紧：跌破 125–127 就要减仓，避免被一轮大级别调整套死。

给我的阿里巴巴港股持仓建议，后续观察看看，和我自己的逻辑大概相似！

[yonwu]

Very interesting! I tried this question:

what price will Nvidia hit in January, 2026?

The whole trajectory and result can be found here
https://dr.miromind.ai/share/6ac4c854-da8b-4075-a14d-d8c3e7997471

We can see a large number of tool calls has been performed!

It also gives out a very nice calculation process

and of coure, it gives a fair prediction in the end, very nice!

[anhui246]

你好，我最近正在休假中，无法亲自回复你的邮件。我将在假期结束后，尽快给你回复。

[karminski]

So I came across this interesting quote from Elon Musk in his Moonshots interview on Jan 6th — he said the ultimate bottleneck for AI will be transformers. And no, not the neural network architecture, he's talking about actual electrical transformers lol.

I got curious and wanted to verify if this holds up, plus check out if there are any investment opportunities in the power and transformer industry. Doing all the research manually sounded exhausting, so I just ran two DeepResearch tasks on MiroThinker.

The prompts were pretty straightforward. First one:

Please help me research the comparison of US-China power capacity in 2025 and AI industry power demand in 2026

Then the follow-up:

Based on the above data, please collect data and continue to analyze and predict the growth of the power equipment manufacturing industry in 2026, including large transformers, power distribution and power management equipment, grid infrastructure and UHV equipment.

It went crazy searching and ended up filtering through nearly 700 articles, then compiled a report with 30+ citations.

Here are some numbers that really stood out to me:

On the power side, China's AI-related power consumption in 2026 is around 100 TWh, while the US is about 80 TWh. Sounds like China uses more, right? But actually the US is under way more pressure — AI accounts for about 2% of total US power consumption, while China's only around 1%. Plus China's installed capacity is almost 3x that of the US, and they've got the whole "East Data West Computing" distributed setup going on, so their ceiling is much higher.

(And I makd a cheat sheet)

The transformer situation is even crazier. Global orders are booked out past 2026, and some customers are willing to pay a 20% premium just to get their stuff. The US imports 80% of its large transformers, and lead times have stretched from 50 weeks to 115-130 weeks... Meanwhile China controls about 60% of global capacity, with 2025 export growth at 47-51%, and leading companies expected to see 15-25% revenue/profit growth in 2026.

So yeah, Musk wasn't wrong — transformers really are becoming a bottleneck for AI infrastructure.

Also gotta shout out MiroThinker's sandbox computing feature. For the numerical predictions, it actually runs Python code to calculate stuff instead of just making things up. Previously tested v1.0 and it used 12.7M tokens with 202 Tool Calls for a single task. This v1.5 version feels noticeably more efficient.

If you've got similar deep research needs, give it a try~

[threeaus]

一看这个 Agent 模型不靠记，主要擅长通过搜来解决问题，我就想起了爱因斯坦，但是野史中这个信息也很多，导致信息很乱，尝试用 mirothinker 来解决这个问题，果然很严谨，有点学究气的感觉，就是这个味儿。