Philipp Kandal

cpo @ grab, before: engineering vp @ telenav, founder skobbler (sold to telenav).

[Philipp Kandal @apphil]

Nikhyl's new Skip episode on PM reinvention is excellent.

His framing: the shift from capital-M Manager to builder is hardest for senior PMs, because what got us here no longer differentiates.

My honest reaction: I'm more excited about being in product than I've been in years. We get to build again.

https://t.co/CIexdTEYwN

[Philipp Kandal @apphil]

Super cool vision how user interfaces of the future can be generated by LLMs on the fly. Mindblowing demo!

[Zain Shah @zan2434]

Imagine every pixel on your screen, streamed live directly from a model. No HTML, no layout engine, no code. Just exactly what you want to see.

@eddiejiao_obj, @drewocarr and I built a prototype to see how this could actually work, and set out to make it real. We're calling it Flipbook. (1/5)

[Philipp Kandal @apphil]

Been waiting for this one. Clear open-source coding SOTA. Would love K2.6 on @cerebras at 1000+ tps and this would be a game-changer. Congrats to our friends @ Moonshot team.

[Kimi.ai @Kimi_Moonshot]

Meet Kimi K2.6: Advancing Open-Source Coding

🔹Open-source SOTA on HLE w/ tools (54.0), SWE-Bench Pro (58.6), SWE-bench Multilingual (76.7), BrowseComp (83.2), Toolathlon (50.0), Charxiv w/ python(86.7), Math Vision w/ python (93.2)

What's new:
🔹Long-horizon coding - 4,000+ tool calls, over 12 hours of continuous execution, with generalization across languages (Rust, Go, Python) and tasks (frontend, devops, perf optimization).
🔹Motion-rich frontend - Videos in hero sections, WebGL shaders, GSAP + Framer Motion, Three.js 3D.
🔹Agent Swarms, elevated - 300 parallel sub-agents × 4,000 steps per run (up from K2.5's 100 / 1,500). One prompt, 100+ files.
🔹Proactive Agents - K2.6 model powers OpenClaw, Hermes Agent, etc for 24/7 autonomous ops.
🔹Claw Groups (research preview) - bring your own agents, command your friends', bots & humans in the loop.
-
K2.6 is now live on https://t.co/YutVbwktG0 in chat mode and agent mode.
For production-grade coding, pair K2.6 with Kimi Code: https://t.co/uvoSJKyGCY
-
🔗 API: https://t.co/EOZkbOwCN4
🔗 Tech blog: https://t.co/9wWvgIQSS3
🔗 Weights & code: https://t.co/Be0hjs2RTP

[Philipp Kandal @apphil]

@swyx @aiDotEngineer So cool you are organizing sg edition. If you have time come by our office for a coffee. Also: Plenty of grab folks going. Would love to join the sg tour :)

[Philipp Kandal @apphil]

What stands out to me is the jump in intelligence, accuracy, and function calling - and that they enable search grounding. This is the kind of progress that could make voice agents genuinely useful, not just impressive in demos.

[Logan Kilpatrick @OfficialLoganK]

Introducing Gemini 3.1 Flash Live, our new realtime model to build voice and vision agents!!

We have spent more than a year improving the model + infra + experience, the results? A step function improvement in quality, reliability, and latency.

[Philipp Kandal @apphil]

Google just dropped Gemini 3.1 Flash Live and it looks like a real step forward for real-time audio models. This is super cool as I got a chance to play a bit with it.

[Philipp Kandal @apphil]

@MihaiSerban Yeah. Also latest Google embedding models fully multimodal…

[Philipp Kandal @apphil]

World Models are the next evolution after LLMs, and the most exciting thing happening in maps right now.

The Seoul World Model from Naver/KAIST: a real, promptable Seoul built from 1.2M street-view images. Navigate freely for kilometers.
Reshape scenes with text. No hallucinated cities, this is grounded in reality.
Incredible work https://t.co/SpmTHqo2kp

[Philipp Kandal @apphil]

@EntireHQ 0b1 (Obi) because every Repo will need a Jedi Master AI to help them :)

[Philipp Kandal @apphil]

Congrats @ashtom and entire team to the launch! Super exciting that we’re getting an ai native dev platform.

[Entire @EntireHQ]

Beep, boop. Come in, rebels. We’ve raised a 60m seed round to build the next developer platform. Open. Scalable. Independent. And we ship our first OSS release today.
https://t.co/OvPKCcjXbq

[Philipp Kandal @apphil]

@MihaiSerban Right in time to switch to opus 4.6 :)

[FailSafe @getfailsafe]

What happens when AI agents can transact, coordinate compute, and operate onchain at scale?

Across Asia, this infrastructure is already live. Agent-native payments, decentralized GPU networks, and onchain identity systems are moving into production at blazing speeds.

Today we’re releasing The 2026 AI × Blockchain Convergence Report, built with @Superscrypt, @base, and @awscloud, with contributions from @GrabSG , @AethirCloud, @virtuals_io, @MessariCrypto, @chatandbuild, @Sogni_Protocol, @KaitoAI and others.

LFG!

[Philipp Kandal @apphil]

Sora 2 is mind-blowing. Good luck @sama mining for more GPUs

[Philipp Kandal @apphil]

@AlirezaGhods2 @solana Thanks so much for organizing / hosting - so great to see all the innovation you are driving . Really happy for us partnering together!

[Philipp Kandal @apphil]

Just in time for Golden Week: Tencent’s HunyuanImage 3.0 is open‑source. 80B params MoE, state of the art performance level. Massive kudos!

[Tencent Hy @TencentHunyuan]

We’re excited to announce the release and open-source of HunyuanImage 3.0 — the largest and most powerful open-source text-to-image model to date, with over 80 billion total parameters, of which 13 billion are activated per token during inference.The effect is completely comparable to the industry’s flagship closed-source model.🚀🚀🚀

HunyuanImage 3.0 originates from our internally developed native multimodal large language model, with fine-tuning and post-training focused on text-to-image generation. This unique foundation gives the model a powerful set of capabilities:

✅Reason with world knowledge
✅Understand complex, thousand-word prompts
✅Generate precise text within images

Different from traditional DiT architecture image generation models, HunyuanImage 3.0’s MoE architecture uses a Transfusion-based approach to deeply couple Diffusion and LLM training for a single, powerful system. Built on Hunyuan-A13B, HunyuanImage 3.0 was trained on a massive dataset: 5 billion image-text pairs, video frames, interleaved image-text data, and 6 trillion tokens of text corpora. This hybrid training across multimodal generation, understanding, and LLM capabilities allows the model to seamlessly integrate multiple tasks.

Whether you're an illustrator, designer, or creator, this is built to slash your workflow from hours to minutes. HunyuanImage 3.0 can generate intricate text, detailed comics, expressive emojis, and lively, engaging illustrations for educational content.

The current release focuses solely on text-to-image generation and future updates will include image-to-image, image editing, multi-turn interaction, and more.

👉🏻Try it now: https://t.co/hHhQpNxJ4e
🔗GitHub: https://t.co/B2bkhlbXxk
🤗Hugging Face: https://t.co/s0G2Lix3ZD

[Philipp Kandal @apphil]

@FanaHOVA Maybe haiku 4.5 - that would be great

[Philipp Kandal @apphil]

@AjeyGore Openai dev day next week sounds like a good opportunity to meet those folks :)

[Philipp Kandal @apphil]

Meituan just dropped a new LLM with “Dynamic Activation.”
Think of it as a brain that decides when to think harder: it activates more experts for tricky parts of a question, fewer for easy ones.
Closer and closer to how our own brains allocate effort. 🧠⚡

[Meituan LongCat @Meituan_LongCat]

🚀 LongCat-Flash-Chat Launches!

▫️ 560B Total Params | 18.6B-31.3B Dynamic Activation
▫️ Trained on 20T Tokens | 100+ tokens/sec Inference
▫️ High Performance: TerminalBench 39.5 | τ²-Bench 67.7

🔗 Model: https://t.co/Ljsrj9sI2H
💻 Try Now: https://t.co/Bfu4mIi8fH

[Philipp Kandal @apphil]

@sama Much faster speed. Run it on groq or cerebras chips. Priority processing good first step- but make it even faster please

[Philipp Kandal @apphil]

Qwen is really on a roll lately. Image generation model looks amazing with text generation and ultra high accuracy rendering. Thanks for releasing it under an open license!

[Qwen @Alibaba_Qwen]

🚀 Meet Qwen-Image — a 20B MMDiT model for next-gen text-to-image generation. Especially strong at creating stunning graphic posters with native text. Now open-source.

🔍 Key Highlights:
🔹 SOTA text rendering — rivals GPT-4o in English, best-in-class for Chinese
🔹 In-pixel text generation — no overlays, fully integrated
🔹 Bilingual support, diverse fonts, complex layouts

🎨 Also excels at general image generation — from photorealistic to anime, impressionist to minimalist. A true creative powerhouse.

Blog:https://t.co/FZ415nwcAp
Hugging Face：https://t.co/XSLwjJ6lYa
ModelScope：https://t.co/mcZ0dHeD64
Github：https://t.co/A9yvJZ6TJc
Technical report：https://t.co/tld83OrOUP
Demo: https://t.co/MgEAr4uqoR