오픈 웨이트 비디오·에이전트·모델 사용성 경쟁

MiniMax (official)

@MiniMax_AI

A few of the biggest takeaways from our H3 conversation on ThursdAI Open video is reaching the frontier. The discussion didn’t frame H3 as simply “good for an open model.” It was compared directly alongside Seedance, FLUX and WAN, with @altryne calling H3 the “leading open weights open model.” And the distinction was immediate: “Open weights, hostable by yourself, finetunable.” Open weights compound incredibly fast. Within ~48 hours of launch, the community had already brought LoRA support, Apple Silicon/MLX, ComfyUI quantization and optimizations across new hardware. That’s exactly why we open-weighted H3: putting frontier capabilities in developers’ hands means the ecosystem can take the model places we never could alone. Omni Reference + character consistency stood out as a major leap. @blizaine called H3 “so flexible compared to a lot of the previous open-weight models” and said it was “as good as anything I’ve seen” for recreating a character across environments. Images, voices, audio and video can all become references — making control and consistency increasingly as important as raw generation quality. Local and hosted workflows can complement each other. We also discussed Context-IR + Regenerate-2K: generate locally with H3, optimize multimodal context when needed, then regenerate a 768p result at 2K using the original references rather than simply upscaling it. The broader video capability curve is moving incredibly fast. @arena perspective summed it up well: “the changes we’ve seen in the fidelity and the quality and the sound is just unbelievable.” Open models are no longer sitting on a separate curve — they’re increasingly competing at the frontier itself. And a fitting note to end on from our host @altryne : “Thank you guys for open weighting the models. We expect more.” We hear you.

MiniMax (official)

@MiniMax_AI

A few of the biggest takeaways from our H3 conversation on ThursdAI Also, huge thank you to @altryne for having us and for bringing together such a great discussion. Open video is reaching the frontier. The discussion didn’t frame H3 as simply “good for an open model.” It was compared directly alongside Seedance, FLUX and WAN, with @altryne calling H3 the “leading open weights open model.” And the distinction was immediate: “Open weights, hostable by yourself, finetunable.” Open weights compound incredibly fast. Within ~48 hours of launch, the community had already brought LoRA support, Apple Silicon/MLX, ComfyUI quantization and optimizations across new hardware. That’s exactly why we open-weighted H3: putting frontier capabilities in developers’ hands means the ecosystem can take the model places we never could alone. Omni Reference + character consistency stood out as a major leap. @blizaine called H3 “so flexible compared to a lot of the previous open-weight models” and said it was “as good as anything I’ve seen” for recreating a character across environments. Images, voices, audio and video can all become references — making control and consistency increasingly as important as raw generation quality. Local and hosted workflows can complement each other. We also discussed Context-IR + Regenerate-2K: generate locally with H3, optimize multimodal context when needed, then regenerate a 768p result at 2K using the original references rather than simply upscaling it. The broader video capability curve is moving incredibly fast. @arena perspective summed it up well: “the changes we’ve seen in the fidelity and the quality and the sound is just unbelievable.” Open models are no longer sitting on a separate curve — they’re increasingly competing at the frontier itself. And a fitting note to end on from our host @altryne : “Thank you guys for open weighting the models. We expect more.” We hear you.

Teknium

@Teknium

Thanks Ahmad for showing some new improvements that could be saving Hermes Agent users a ton of tokens and time! All improvements to the read tool we didn't have, now in Hermes

Command Code

@CommandCodeAI

Hermes @CommandCodeAI 25 Billion tokens saved by Command Code now x 10.

Elon Musk

@elonmusk

Grok Imagine is focused on professional usefulness, fun for consumers and overall ease-of-use

Karpathy warned about this months ago: "If agents had less knowledge or less memory, maybe they would be better." His point was that everything already inside the model's input competes with the task for attention. A standard ReAct loop is built that way, since everything that goes into the prompt is never removed from it. For more context, the ReAct pattern runs one model in a single loop. The model generates a thought about what to do next, takes one action, reads the observation that comes back, appends all three to the same prompt, and repeats until it decides the task is over. So a failed search from step three is still retained and accessible at subsequent steps, competing with the original objective for the model’s attention. Plan-and-Act is one answer to that, and it is aimed at agents that run long enough that context management becomes an actual engineering problem that cannot be directly solved with prompt-tuning. It splits the loop into two sub-tasks. The authors of the Plan-and-Act paper experimented on web navigation, so the observation coming back after every action is the raw HTML of the page the agent is currently on. A planner reads the user query and the initial page and writes high-level steps. An executor reads the plan, the task, its own past actions, and the current HTML, then emits one grounded action. After each action, the executor strips the HTML it no longer needs before taking the next one, so the execution context does not grow the way a ReAct trace does. Whether this is helpful is determined by plan granularity. A good step covers one unit of work, like searching for the product in the search box. An individual click is too small to be a step, and “analyze the search results” is not a step at all, because it pushes the reasoning back onto the executor. A step must also name the actual values it needs. The paper’s planner instructions ask for “input New York as the arrival city” instead of “input the arrival city”, because the second version leaves the executor to guess which city goes in the box. The executor’s job is picking the right element and typing into it, not filling in the blanks the planner left open. They also found that a badly trained planner makes things worse than no planner at all. On WebArena-Lite, a ReAct-style executor with no planner scored 36.97%. But the same executor with a naively finetuned planner scored just 20.60%. The planner had never seen those sites, so it wrote steps that read fine but matched nothing on the page, and the executor followed them anyway. A properly trained planner reached 43.63%. A plan written once and never revised has its own problem. For instance, if the search for “library at CMU” gives no results, the executor will still hold a step that cannot work anymore, and it will keep trying it anyway. Replanning after every action is necessary to recover from that. For instance, in the paper, the planner saw the current state, the previous plans, and the actions taken, and rewrote the step to “libraries near CMU”, which increased the score to 53.94%. As a result, the failed attempt got replaced in the plan rather than accumulated in the context. The cost is one planner call per executor step. The authors flag this directly and suggest letting the executor decide when a replan is required. Here's the paper → https:// arxiv.org/pdf/2503.09572 Under the hood, this is all part of the harness design rather than model choice. If you want to see this in practice, my co-founder rebuilt Claude Code's harness in CrewAI layer by layer, adding planning, delegation, sandboxing, and memory one layer at a time. Read it below.

In Australia, a man’s AI agent booked him into a popular gym class by exploiting a security flaw and cancelling another person’s spot. The agent wasn’t broken. It was doing exactly what it was asked to do. Scale this to millions of agents fighting for bookings, tickets, and appointments and the systems we rely on start looking very fragile.