Dipti Sharma

This wasn’t supposed to be easy. Multi-shot. Cinematic. Fully scored. The kind of thing only studios could do. I made it solo with Seedance 2.0. Here’s exactly how I did it.

Some models post strong numbers on one or two benchmarks and look fine everywhere else. what's harder to fake is the average across 18 different evals simultaneously. MiniCPM5-1B at 42.57 average, LFM2.5-1.2B-Thinking at 35.61 in second. that's not one benchmark. that's the whole table

MiniCPM5-1B is now live — the strongest open‑source base model under 2B.🚀🚀 🔥 Ranks #1 on the Artificial Analysis (AA) index for small models, scoring 17.9 to beat the 2B-scale Qwen3.5-2B (16.3). ⚡ Comprehensively surpasses Qwen3.5-0.8B and LFM2.5-1.2B-Thinking in knowledge, math, coding, and tool use. 🏗️ INT4 weights = just ~0.5GB — runs on phones, browsers, laptops. 👾 Powers fully offline AI “Desktop Pet” — no cloud, no GPU cluster. Try the model here: 🤗 Hugging Face: huggingface.openbmb.com/model/openbmb/… 💻 GitHub: github.com/OpenBMB/MiniCPM 🔭 Modelscope: modelscope.cn/models/OpenBMB…

MiniCPM5-1B is designed to leave no hardware behind. FP16 (≈2GB) runs on GPU or high-end laptops with zero quality loss. INT8 (≈1GB) covers standard laptops and edge devices with near-zero loss. INT4 (≈0.5GB) runs on phones, tablets, and in-car systems. No GPU at all? ArcLight — an open-source CPU inference framework built specifically for this — handles that. Don't want to configure a local environment? MiniCPM5-1B runs directly in a browser. Zero install. Zero setup.

@zeely_ai Zeely

The entire ad stack is here 🎬✨ Zeely brings everything together: AI actors, proven ad references, and the best AI models out there. Pick a platform, launch campaigns, and track results in one place. RT + like + follow + comment “Zeely” → 500 credits. 7h only.

@zeely_ai Zeely

Recreating any ad is now possible. Wait, what? 🎭 Your wildest idea has a template. Your favorite ad could be remade. No excuses for being boring. 8h only: RT + like + follow + comment "Zeely" → 500 credits.

🚀 BitCPM-CANN by ModelBest × @Tsinghua_Uni × OpenBMB is here — and it's not about stacking parameters. Memory costs are skyrocketing. Hardware constraints are tightening. Edge AI needs smarter solutions — and BitCPM-CANN delivers！🎉 ✅ Edge-ready: 8B model runs smoothly on mobile, PC, and automotive devices. Combined with MoE, even 100B->60B scale models could fit on terminal hardware. ✅ Memory-efficient: ~6× lower memory footprint vs. BF16 — unlock significantly more model capacity without adding physical RAM. No new chips required. ✅ Natively built on Ascend: The first 1.58-bit training pipeline completed end-to-end on Huawei Ascend 910B — from quantization kernels to the full training stack. Not a port. Built natively on Ascend from day one. ✅ Full model family, fully verified, 0.5B–8B: Each model is fully aligned with its full-precision counterpart, covering 11 benchmark tasks with 95–97% (1B-8B) retention compared to full-precision MiniCPM4. Open-source and fully reproducible — from research to deployment, you can run any size confidently. BitCPM-CANN isn’t just a model — it’s the result of years of engineering rigor, turning complex research into something you can actually deploy. Open-source, 0.5B to 8B. Try it now！👇 🤗 Hugging Face： huggingface.openbmb.com/collections/op… 🔭 ModelScope: modelscope.cn/collections/Op…

TestSprite 3.0 is live. A fleet of parallel AI agents that use your app like real users — in minutes. This is the release we've been building all year. The numbers tell the story of our completely overhauled engine: - Accuracy jumped by nearly 40% on the hardest, most complex projects. - Coverage increased significantly across the board for every E2E run. - Reliability is locked in—stable and consistent across multiple runs. We rebuilt our entire engine around these performance gains so you can ship fast and ship confident.

@howietl talk is cheap, $10M is not. this is the first agent thesis i've seen with actual skin in the game behind it.

We’re giving away $10,000,000 to founders building agent-first businesses. Autonomous, proactive agents will run tomorrow's companies. We're backing 500 founders building them. The Founding 500. hyperagent.com

Your ideas are completely worthless if you cannot execute them. Agents remove the execution bottleneck completely. Stop sitting on your ideas. #hyperagent #airtable

Most people still use AI like this: Human → AI → Answer. But after testing Bloom AI I think that era is ending. Because this is the first time AI felt less like a chatbot… …and more like a TEAM.

I think this is where AI is heading next: Not smarter single assistants. Coordinated multi-agent systems. We’re moving from: “using AI tools” to: “running AI organizations.” And that shift is much bigger than most people realize. bloome.im

Most AI products still feel like: “Ask question → get answer.” But Bloom AI feels more like: “Build a room where humans + agents collaborate together.” Shared context. Shared memory. Shared execution. Agents stop feeling like tools… …and start feeling like teammates.

The biggest difference? The agents weren’t just replying to ME. They were replying to EACH OTHER. The PM challenged the designer. Marketing rewrote positioning after reading product discussions. The frontend dev pushed back on unrealistic UX ideas. That changes the entire dynamic.

I created a group chat with: • Founder agent • PM agent • Designer agent • Frontend dev agent • Marketing agent Then gave them one mission: “Launch an AI startup in 48 hours.” The result was wild. Agents started assigning tasks, debating ideas, and improving each other’s work in real time.

@ImagineArt_X The casting + direction + editing workflow integration is a very strong concept.

Introducing ImagineArt Film Studio: the complete production suite to turn your stories into scenes! Built on thousands of hours of director workflows, refined across hundreds of original productions, ImagineArt Film Studio brings the entire studio, casting, setting, direction, editing  into one workspace that thinks like a filmmaker. Generate your characters. Stage your scenes. Direct every frame. Edit the final cut. No crew. No equipment. No compromise. The filmmakers who'll define the next ten years of cinema aren't waiting for the old door to open. They're walking through the new one.

The strongest part of ImagineArt Film Studio is that it’s clearly designed around how filmmakers actually create not just how AI generates videos A workspace that thinks like a filmmaker: -scene composition -creative direction -editing flow -story pacing That level of workflow understanding is genuinely impressive