indie dev & product thinker (INTJ) building 「JustLog APP」 &「Dino-GSP Web」 exploring how systems & creations grow with warmth.

Hangzhou China
Simple truths are always the hardest to act upon.
If you get addicted to things easily, you have one of the rarest traits there is. Most people can't stay interested in anything for more than a week. You can lose 6 hours to something without noticing. Right now that goes to games, feeds, and whatever hooks you first. That's the only problem. Pick one hard thing actually worth being addicted to. A skill, your body, your work. Give it the same obsession for 2 years and you'll pass people who've been at it for 10.
7
Luhui retweeted
Energy capacity = economic capacity, in one chart There are no low electricity, rich countries
"Energy Capacity = Economic Capacity" Former Intel CEO Pat Gelsinger on why America's power grid is the ceiling on growth: "We went through 10 to 15 years where essentially I was taking coal offline at the rate I was adding renewables, and essentially the nation was flatlined in terms of overall energy capacity." "So essentially our economic capacity as a nation, flat for fifteen years." "Even with that, we've gone to maybe increasing our national energy capacity 4% per year. Going from essentially zero to one." "Nuclear as base load is a fabulous thing that we want to go build more of. Unfortunately all of our renewables have deep dependencies in China." "Energy capacity somewhat dampens how euphoric we can get on AI. Why build a new data center and buy the million GPUs if I can't power them?" @PGelsinger @RaghuRaghuram @appenz
277
906
4,656
3,691,658
Luhui retweeted
据说是 AI 初创公司目前最想要又最难招聘的职位 > 长期在线(了解趋势) > 有品味且能创作(新媒体) > 理解 AI 和技术 > 能高质量执行(执行者) 如果按照技术理解力和内容能力分成四个象限画张图的话,图里能看出为什么这个岗位难招: 现有的几类人各自只占一半,内容创作者和社媒运营在左上,有网感但讲不清 AI 产品;技术布道和工程师在右下,懂技术但内容没人看;传统公关执行力强,却既不够“在线”也不够懂技术。 右上角要的是四项同时具备的人,这类人现在本来就少。 不知道我能不能满足这要求?
literal hardest role to hire for rn and every ai startup wants someone > chronically online (knows trends) > has taste & can create (new media) > understands ai & technical > can execute with quality (operator) such a small niche of folks who can do this
196
68
590
125,070
Luhui retweeted
Gesture recognition using the Decisions API by @OpenAIDevs. Thanks to it's vision understanding it is way more accurate than my previous Jev-based demo. It was also faster. cc @thsottiaux @jxnlco @pvncher @tldraw @steveruizok
Jev as a shape recognizer. Results are mixed, but damn it’s fast. Multi modal coming though (See comments), hopefully that “just works”
46
78
1,298
107,059
Luhui retweeted
Mathematicians are coping so hard. Suddenly they can not stand AI. 😀 the guild is so badly protecting its turf. The Association for Human Mathematics (AHM), criticised OpenAI's yesterday's release of 722 math manuscripts and urged mathematicians "to discontinue their work with OpenAI". And then Terence Tao's reposted AHM's statement on his blog. Walking away is not a scientific position.
187
132
1,390
167,045
Why are mathematical problems still hard for AI Agents? The challenge is not only generating proofs. Agents must also choose useful strategies, manage dependencies, prevent errors in long reasoning chains, and preserve insights from failed attempts. luhuidev.medium.com/da57ec19…
1
10
Reducing LLM costs isn’t mainly about making prompts shorter. The bigger savings often come from eliminating repeated context, oversized tool outputs, ineffective retries, unnecessary model escalation, and agent loops that should have stopped earlier. luhuidev.medium.com/6b4591ab…
1
9
Luhui retweeted
since you guys loved the exploding tesla.. I used GPT-6 Astra to create a 3D website that pulls apart the male anatomy into 2,234 modeled pieces! we are in a renaissance of learning
1,152
5,422
53,372
8,033,411
More agents ≠ better collaboration. In Multi-Agent systems, I keep seeing the same failures: - roles without real interfaces - “parallel” tasks with hidden dependencies ...... My seven lessons from building Multi-Agent workflows: medium.com/p/6bbbf8c56701
1
13
AI is making execution cheaper. The durable advantage is no longer just “can you code/design/write?” It is whether you can close the loop: discover → specify → model → orchestrate → verify → deliver. A personal skill map for the AI era: luhuidev.medium.com/3f0bdabf…
1
25
CAPTCHAs assumed humans could solve what machines couldn’t. In 2026, that line is collapsing—from OCR and image grids to GUI agents that click, drag, and self-correct. What replaces the puzzle? Risk, identity, and cost. luhuidev.medium.com/a55b439f…
1
4
cry
worktrees must die this is 20GB of repeated node_modules lol
2
AI can generate a convincing research paper—but can it prove that its claims are true? This explores Google ScientistOne’s Chain-of-Evidence framework and why verifiable evidence may be the next essential layer for trustworthy AI agents. Read it here: luhuidev.medium.com/680e7295…
3
13
‼️ BREAKING: An active npm supply chain attack has compromised at least 868 packages carrying over 2 billion monthly installs with a credential-stealing worm. Shai-Hulud is back. It started with the compromise of the GitHub account of the maintainer behind keyv, a library with roughly 127 million weekly npm downloads. A preinstall hook fires on npm install and drops a stealer that sweeps npm, GitHub, AWS, Kubernetes and Vault secrets, and then spreads to more maintainers.
309
1,336
8,669
1,440,125
Luhui retweeted
It’s incredible that you can let Opus 5 run for two hours in Claude Code and get a cinematic Three.js movie inside one HTML file. It created the story, scenes, transitions, camera moves, and procedural world. Higgsfield generated the narration, sound effects, and music. I mostly helped with references for more realistic grass, added the skills I use, and asked it to keep verifying its work. It’s all open source, so you can play with the code. HTML Site: mengto.github.io/a-long-expe… GitHub: github.com/MengTo/a-long-exp… Grass reference: toolcraft.sh/grass
We're starting to leave the territory where you'd test an LLM by e.g. "create an svg of pelican on a bicycle". As one idea to generalize it, I was interested what Opus 5 would do if I gave it the first paragraph of the Lord of the Rings, a 1M token budget (~$10) and asked for three js render of it. Opus went off for ~2 hours and wrote 5500 lines of code that (procedurally) rendered the story. It's kind of janky but fun. But it's a bit mindboggling that the LLM has to place and orchestrate various polygon assets in (x,y,z) coordinates and write code that animates it all, and that it even does anything at all. I also like this kind of examples because no one in their right mind would ever spend the time to write something this custom but LLMs have all the stamina and patience in the world, so it's an example where we go from "no one would ever do this" to "sure, why not, it's ~free". There might be a lot more. But I'm excited about creating hyper custom worlds that you can imagine dropping players into, e.g. here to participate in the LoTR story as a spectator NPC, or one of the characters, or etc. Something like an ephemeral GTA of X on demand. Last thought is that the domain of worlds/games exposes a weakness in LLMs: they can't easily audit their work because they aren't able to efficiently and natively perceive videos or play games within them. Here, Opus 5 had to very slowly and painstakingly take screenshots at different points, and it messed up a few times and created a bunch of jank. An example of raw capability (multimodal, gameplay) that I think is still quite lacking.
17
34
515
69,897
Luhui retweeted
🚀 DeepSeek-V4-Flash Official API is now LIVE in public beta! 🔷 We’ve massively upgraded its Agent capabilities—benchmark scores are now far surpassing the V4-Pro-Preview. Check out the massive performance leap below! 👇 🔷 The official V4-Flash now natively supports the Responses API format and is fully adapted for Codex! Check out the configuration details in our official API docs: api-docs.deepseek.com/quick_…
1,699
3,499
29,860
9,599,904
Loop Engineering can keep an AI agent working—but not necessarily working on the right thing. More runtime and more code don’t guarantee better software. The real investment should be in Harness Engineering: clear boundaries, reliable verification.. 🔗 luhuidev.medium.com/p/71cb65…
1
13
Prompt Engineering optimizes one interaction. Loopcraft optimizes the system that keeps running: triggers, verification, retries, state, and feedback. The next abstraction in AI agent engineering may not be the prompt, but the loop. medium.com/p/9bed5d3ce104
1
15
Your AI support bot gets a ticket needing data from 4 systems, 4 teams, 4 permission models. Traditional RAG: one search, one shot, "I don't know." Agentic RAG: plan → retrieve per-source → check what's missing → retrieve again. medium.com/p/b15711bb1bb2
1
37
Luhui retweeted
🚨 OpenAI pondering big price cuts, per WSJ scoop:
17
36
171
11,314