Head of R&D at MiniMax Code. Co-creator of DeerFlow. Building the next generation of AI agent systems.

北京
hello, world! i'm daniel! currently i'm the head of r&d of MiniMax code @MiniMax_AI, which is a frontier ai lab in china. i enjoy working on our ai-native product @MiniMaxAgent . i'm really obsessed with this question: what does an ai-native organization look like? how can our organization co-evolve together with llms? besides that, i'm passionate about open source. i'm the co-author of the deerflow project, which is a representative work in harness engineering and has 78k 🌟 on github. if you're passionate about open source, foundation models, or agents- or if you enjoy playing basketball or riding a road bike - let's chat and be friends! :)
65
18
311
179,000
RonnyHe retweeted
I put Anthropic's whole harness playbook for long-running coding agents on one page honestly the most useful thing I've read on agents all year two Anthropic engineering posts + the Opus 5.5 launch, squeezed into one sheet: > a planner grows a one-line idea into a 16-feature spec > a generator builds it, an evaluator opens the app in Playwright and fails any sprint below the bar > session one writes init.sh, a progress log and 200+ failing features. every session after closes one and commits > the 4 ways agents wreck long tasks, and the fix for each > 3 prompts to copy: coding agent, QA agent, planner solo run: 20 minutes, $9, the game didn't even respond to input full harness: 6 hours, $200, playable QA was about 8% of the bill on their DAW run and Opus 5.5 runs about 40% cheaper than Opus 5 bookmark this before you wire up your next agent
19
7
61
2,175
RonnyHe retweeted
"control, we have touchdown"🥕🐰 landing confirmed! M3.1 Flash by H3
Welcome back. Space Bunny. 🚀 MiniMax-M3.1-Flash is coming.
2
1
43
1,471
If you've played with Claude Code mods or DeepSeek Harness plugins, it's the same vibe: bend the agent to fit how you work. Except here you just tell the Agent what you want and it builds the whole app for you.
2
4
616
Seriously, we'd love everyone to jump in. Build the tools you wish existed, share them, and let's grow this ecosystem together 🚀 A few of my friends are already gearing up to contribute: @henry19840301 @SToneoneX @jasonzhou1993 @treg_ai @crewlet_
3
389
PowerContext is an Open Source context-layer project, genuinely glad to see this collaboration.
MiniMax Code now works with PowerContext by @OceanBaseDB Your agent remembers project decisions with their sources, and hands off unfinished work so you can pick it up in one click, whether that's tomorrow, in a new session, or by a teammate. Available now in the MiniMax Code Plugin Marketplace.@dinah_zhang github.com/oceanbase/powerco…
1
23
1,132
RonnyHe retweeted
Two offers, now live in MiniMax Code: - Unlimited M3.1 Flash Through Oct 25, 23:59 (UTC+8), every MiniMax Code subscriber can use M3.1 Flash with no usage caps. - 50% off your first month of M Plan Through Oct 31, 23:59 (UTC+8), new subscribers and M Plan upgraders get 50% off their first month. Standard pricing applies from month two.
Introducing MiniMax M3.1 Flash, our latest model, built for real work. An everyday workhorse for coding, research, and creative tasks. Now available in MiniMax Code.
29
19
213
13,361
Welcome back on the App Store, MCode!
You've been asking for this one... MiniMax Code is back on the App Store.
5
54
2,597
Intelligence with everyone! Let's welcome M3.1 Flash and wait for the Pro model.
47
16
493
32,396
MiniMax M3.1 Flash is Coming. And it will be free in MiniMax Code for every subscriber through Oct 25. No credit burn, no limits. Let's ship. 🚀
47
38
507
27,796
MiniMax-M3.1-Flash is Coming! Stay tuned!
Welcome back. Space Bunny. 🚀 MiniMax-M3.1-Flash is coming.
10
4
116
4,446
Space Bunny!
Dear friends, we are back.
37
1,384
AI is changing how needs get met. Before: a few PMs decided what every user needed. Now: a team builds the infrastructure, atomic tools, a great agent, and a way to extend it. Then users and their agents build what they need themselves. Mini Apps in MiniMax Code are built on this idea. We provide a few basics (a port, an agent, your data) and draw clear boundaries around each. What gets built on top is up to you and your agent. Need a small tool of your own? Just ask the agent. From "we decide for you" to "we give you the power to build it." That's the real magic of this era.
17
2
55
3,418
We built the stage. You create. MiniMax Code now has MiniApp: a way to make MCode grow into something that's yours. The first one is MiniMax Bot, built by @MiniMaxAgent and me: · A chief bot that builds your team, opens group chats, sets up scheduled tasks · @ hands off in a group chat; bots can pass work to each other · Heavy jobs get moved to the background — you keep talking, it comes back with the result · Every bot has its own model, reasoning effort, memory, and look We ship the stage. You decide the play. #MiniMaxCode #MiniApp
44
4
113
36,923
It's currently in invite-only beta and will be officially released later this week. You won't need to download a separate app—just update your MiniMax Code client to the latest version, and MiniMax Bot will appear in the Plugins section.
3
944
RonnyHe retweeted
Writing code is easy now. You can build an app just by saying what you want. But things are very different in production. I'm a hands-on developer, and writing code may account for less than 10% of my total delivery time. Most of the rest goes into review and verification. No matter how far models advance, I never seem able to fully let go. I always have to run their code myself to see whether it meets the requirements correctly. This is the main bottleneck keeping me from delivering at scale. So the way I see it, the key to large-scale parallel delivery is letting models verify the code they write, and making that verification something we can trust. If I could keep only one of the Skills I've written over these two months, I'd keep the one that builds the verification environment: the agent starts the app itself, uses it like a user, then checks the results against a checklist organized by feature, with an ID for every criterion. I call it a feature map. How I built it and the problems I ran into are all in this article.
In August, I built my own multi-agent orchestrator to have agents work in parallel and review each other’s work. By late September, I’d abandoned it. This is the full retrospective: why I abandoned it, how I work now
Article

Why I abandoned my own multi-agent orchestrator

From multi-agent orchestration to verifiable delivery: lessons from two months of AI-assisted development. Here is the one thing I'm most sure of after the past two months: Scaling agent work and

2
3
7
482
After weeks of living inside Grok Bot, I'm convinced the next step is combining that assistant experience with coding agents. So I'm building something fun. More soon.
11
64
5,889