Every week, I organize 400 to 700 links into roughly 60 categories as part of my ongoing effort to learn about AI. This is my personal notebook, which I enjoy sharing with friends… a hobby and a labor of love, rather than a commercial publication or product.
If you arrived here through a search or shared link, this page collects the links I found for International for the week ending August 21, 2026.
As part of my learning process, I like to automate the category covers. It gives me a chance to learn Python and APIs.
This week’s cover prompt was written using Claude Opus 4.7, and the image was generated using Nano Banana 2.1.
Category Cover Image Prompt: Bright sunlit architectural photograph of a 1929 Art Deco limestone skyscraper setback where four identical stylized herms, carved from the same honey-cream Indiana limestone as the wall and emerging from narrowing piers between dark steel-framed windows, together cradle one shared carved stone globe incised with Deco meridians and simplified continents. A frieze band directly beneath the herms is deeply incised with the words INTERNATIONAL in Art Deco capitals with crisp shadowed cuts, the facade raked by angled sunlight, with neighboring 1930s buildings and a deep blue sky receding behind, perspective-corrected verticals, realistic architectural photography.
This Week in International News
Summary by Claude Sonnet 5.5 based on this week’s links:
- Qwen3.8-27B lands as a local model: Most of the week's buzz was about Alibaba's Qwen3.8-27B running on a single consumer GPU. Cline reported it became the top local model in four days, and Artificial Analysis scored it near much larger models. Skeptics pushed back: Mollick said it falls well short on complex agentic tasks, and one user said it failed a one-shot fluid simulation prompt.
- GLM-5.3 gains come from post-training: Z.ai launched GLM-5.3 at the same API price as GLM-5.2 and Artificial Analysis scored it 60, level with Kimi K3. A Zhihu breakdown argued the gains come largely from post-training, since it reuses GLM-5.2's base model.
- DeepSeek open-sources its agent harness: DeepSeek released its Harness as open source on Aug 13, alongside V4 Pro. It is the layer that turns a model into an agent, with tools, memory and an execution loop. Commentators called it a possible new battleground beyond the models themselves.
This summary was generated by Claude Sonnet 5.5 to help you explore the links below. Rest assured, I select, organize, and check the links by hand in Google Sheets, and write the introduction and personal commentary in The Main Newsletters myself each week as a labor of love.
This week's links related to International
Record & Replay is also available in the ChatGPT app for macOS across the EEA, UK, and Switzerland. Turn your go-to workflows into reusable skills by showing, not just telling, ChatGPT Work and Codex what to do.”
https://x.com/OpenAIDevs/status/2090487779587477626
DeepSeek is having its second DeepSeek moment It’s opening up the layer many thought would be the next moat – the harness. Their DeepSeek Harness is the layer that turns a model into an agent: tools, memory, execution loop, sandbox, storage, scheduling, interface. (DeepSeek has”
https://x.com/TheTuringPost/status/2089661979287835076
🔓 DeepSeek Harness Looks Half-Baked At First Glance … Then You Realize That’s The Whole Point DeepSeek Harness went fully open-source on Aug 13, the same day V4 Pro shipped and API pricing jumped to peak/off-peak rates … the repo blew past tens of thousands of GitHub stars”
https://x.com/ZhihuFrontier/status/2089998555889250478
Introducing Agentic Search | Mistral
https://mistral.ai/news/agentic-search/
Exciting news: Dreamina Seedance-2.5 by @BytePlusGlobal is topping ranks in Video Arena! It currently ranks #1 in Video Edit, the industry’s newest video capability category, which has only 9 models on the leaderboard so far. #1 Video Edit Arena: 1411 pts – a large 23 pts gap”
https://x.com/arena/status/2089448812159045848
Motion Picture Association reaches agreement with ByteDance over AI guardrails – Los Angeles Times
https://www.latimes.com/entertainment-arts/business/story/2026-08-17/motion-picture-association-reaches-agreement-with-bytedance-over-ai-guardrails
GLM-5.3: Frontier Coding with Emergent Cyber Capabilities
https://z.ai/blog/glm-5.3
Interesting take by the zAI (GLM-Models) founder: AI scaling is not over, but we just focused too much on model size. For years, the industry mostly asked: How many parameters does a model have? But performance also depends on training data, inference compute and post-training.”
https://x.com/kimmonismus/status/2090026799916888080
Qwen 3.8 27B is #6 among open-weight models on the Vals Index v2 and is #1 on Harvey’s Legal Agent Benchmark among open weights.”
https://x.com/ValsAI/status/2089836844842435040
Qwen 27B is really good local model but, when you use it, it is immediately absolutely and obviously nowhere near as good as the other models listed here for agentic tasks, and especially for the kinds of complex tasks that GDPval-AA proports to measure Do your own benchmarking!”
https://x.com/emollick/status/2089914722464223706
Qwen3.8-27B ranks #7 on Artificial Analysis’ Agentic Index at only 27B parameters, a fraction of the size (and cost) of the other six. Train and deploy it all on one platform with the Baseten Loops SDK:”
https://x.com/baseten/status/2089749674265551077
Qwen3.8-27b is surprisingly usable with codex on my rtx5090! ~200k context with nvfp4 – *very* steerable with skills and plugins – quite smart! great for frontend, basic debugging, and image I don’t trust it (yet) to run in yolo mode – but *very* impressed by how capable it is”
https://x.com/rishdotblog/status/2089458516092399889
Watching a team of bots split up a large game project based entirely on their own experience and specialties with little input from me is pretty crazy (and just a bit creepy) to watch. I made a short video on group chats in the Hermes Desktop App Bot Mode: – Qwen Bot mentions”
https://x.com/tonbistudio/status/2089226021749030999
Pareto frontier is live in Agent Arena! Dive into model performance on real-world agentic tasks, compared to the median cost per task. The current models on the Pareto frontier for Agent Arena are: – Claude Opus 5 (High) by @AnthropicAI (+12.34%/ $1.78) – Kimi K3 (Max) by”
https://x.com/arena/status/2090137780932538549
@WanHL7 @aquilesfd Theyre all bigger than that, imo. Opus is ~1.5-2T and Fable is ~5-7T. Sonnet is GLM sized (~600-800b). And haiku is probably ~200-350b. Anthropic is better at making Big models but worse at making small models than OAI.”
https://x.com/jmbollenbacher/status/2089712688096022619
The real benchmark for an AI agent isn’t what it says. It’s what it gets done. AutoClaw, @Zai_org ‘s AI Agent for Work, is built to turn intent into finished output. Give it a complex task. AutoClaw breaks it down, operates tools, works across files, coordinates multiple”
https://x.com/AutoClawAIer/status/2090446256342708724
DeepSeek Harness (dsh) – a very convenient open-source agent harness from @deepseek_ai It turns an LLM into an extensible agent system. → The key idea is that everything is a plugin. DeepSeek Harness makes tools, interfaces, integrations, models and agent modular and”
https://x.com/TheTuringPost/status/2090096803899216151
Great to see so much interest in our post-training work. Want to reshare SAO (Single-Rollout Asynchronous Optimization), a key recent advance in agentic reinforcement learning alongside GLM-5.2. SAO tackles the off-policy and stability challenges of asynchronous RL through: -“
https://x.com/ZixuanLi_/status/2090564295696306436
GLM-5.3 API is now live. – Built for coding, defensive cybersecurity, and long-horizon agentic tasks – Priced the same as GLM-5.2 – Available via the official API and partner model gateways Get started:
https://x.com/Zai_org/status/2089816129011098048
A “refusal-removed” version of Qwen3.8-27B can now run locally on Apple Silicon. Even its creators warn that it can provide malware, fraud and weapons instructions on demand. It was released as an MLX build in 2, 4, 6 and 8-bit versions. The uploader claims its 4/6/8-bit tests”
https://x.com/kimmonismus/status/2089763435865088508
literally just tried to let qwen3.8 27b one shot a fluid simulation webapp it thought for 40k tokens, took around 1 hour to generate and it’s just black and doesn’t work meanwhile Opus 4.5 just oneshots it in a minute and works wonderfully exact same prompt”
https://x.com/scaling01/status/2089784644400976254
Ollama has the best performance for deepseek v4 flash on average. For local only, you can try qwen3.8 that is optimized: Apple Silicon: ollama run qwen3.8:27b-mlx NVIDIA: ollama run qwen3.8:27b”
https://x.com/ollama/status/2089454609765146744
Cartesia’s Sonic 3.6 takes the #1 spot on both the Provider Voice and Controlled Voice Artificial Analysis Speech Arena leaderboards, surpassing Speechify AI’s Simba 3.2 and Alibaba’s Qwen-Audio-3.0-TTS-Plus, with Sonic 3.5 holding #2 on Controlled Voice Sonic 3.6 is the latest”
https://x.com/ArtificialAnlys/status/2089400880688976062
Qwen3.8 vs Qwen3.6 vs Gemma 4: 24GB GPU Test
https://kingy.ai/blog/qwen3-8-27b-vs-qwen3-6-27b-vs-gemma-4-31b/
Artificial Analysis Intelligence Index puts Qwen3.8-27B at DeepSeek V4-Pro and GPT 5.6 Luna performance. This is the first time a local model has scored frontier model capability. We weren’t expecting this pace of local progress anywhere near this soon.”
https://x.com/cline/status/2089425906569977896
Qwen3.8-27B is now the #1 local model in Cline after just 4 days. This ends a *4 month* streak by the previous winner Qwen2.5-Coder-7B which has been the top local model since April.”
https://x.com/cline/status/2089825294677143973
We’re releasing new Qwen3.8-27B GGUFs with 10% higher accuracy. Unsloth Dynamic V3 outperforms others by >10% on Div-300, KLD & more benchmarks. We also release 1-bit quants that retain 77% accuracy. Run on 8GB RAM. Blog:
https://t.co/tHsBexyh2K GGUF:
https://x.com/UnslothAI/status/2090103470015828184
4 days to the top.🏆Thank you to every builder who pushed Qwen3.8-27B to #1 on Cline! @cline”
https://x.com/Alibaba_Qwen/status/2089919106522976337
GLM-5.3 (Max) by @Zai_org has shifted the Pareto frontier for Code Arena: WebDev! GLM-5.3 (Max) would sit at #2 among open models (upon open weights release, and #8 overall) with 1597 pts in the Code Arena: WebDev. At $3.65/M, GLM-5.3 (Max) is on par with Qwen3.8 (Max)”
https://x.com/arena/status/2090581559262798055
Qwen 27B is the “DeepSeek moment” for open source. It matches the closed-source state-of-the-art from just a few months ago and runs on an RTX 5090. Without exaggeration, it’s a game changer.”
https://x.com/kimmonismus/status/2089740575830409700
orcarouter/Qwen3.8-27B-Uncensored-MLX · Hugging Face
https://huggingface.co/orcarouter/Qwen3.8-27B-Uncensored-MLX
turns out Qwen3.8-Max can do instance segmentation you send the image plus a class list and ask for JSON (box + closed polygon per object, both in 0-1000 coordinates) we scale those vertices to pixel space and rasterize them into instance masks”
https://x.com/skalskip92/status/2089422495631687759
Qwen3.8 27B as intelligent as Luna Max yoo”
https://x.com/populartourist/status/2089415259199029345/photo/1
We also released 1-bit Qwen3.8-27B quants which run on 8GB RAM. They retain ~77% of accuracy compared to BF16. We originally didn’t want to release them but were shocked that they worked extremely well on our internal testing!”
https://x.com/danielhanchen/status/2090119165055324518
DFlash 2 is here! Qwen3.8-27B at 70 tok/s on an M5 Max MacBook Pro. ⚡ Up to 4.6× the speed of autoregressive decoding, with the same output. This is the next generation of DFlash, seeded at Z Lab and upgraded at Inco AI. Get one more accepted token on every pass, for free!”
https://x.com/zhijianliu_/status/2089836737132650504
3000000000 downloads! Can you count the zeros at a glance? 😎 Thank you all for the incredible love. Let’s keep growing together! 🌱”
https://x.com/Alibaba_Qwen/status/2088881015855182122
Qwen 3.8 27B is excellent, but it defaults to wildly overthinking things
https://simonwillison.net/2026/Aug/16/qwen-38-27b/
We can now RL large MoEs with 0 train-infer mismatch! And doing so can improve performance (pictured task: teach Qwen3.6-35B-A3B to play Wordle). Everything is open-source and we did a bunch of ablations. 🧵”
https://x.com/PandaAshwinee/status/2089396727048749528
MiniMax H3 image+audio to video feed an audio into h3 instead of generating it. i was testing this flow, got shocked with how good the lipsync and movement is implemented it on the i2v version in diffusers, made a demo for it. what do you think? 🔊 ▶️
https://x.com/multimodalart/status/2089418659370357191
This is super cool! Sota video generation models like our MiniMax H3 understand the movement of a walking character so well, while some people report that latest image generation model still struggles to keep walking motion consistent. MiniMax H3 is here for your game sprites”
https://x.com/MiniMax_AI/status/2089420340728610890
GLM 5.3 is #2 on Terminal Bench, #3 on Legal Bench, and #6 on Skills Bench among-open weight models.”
https://x.com/ValsAI/status/2090192848780136668
GLM-5.3 achieves 60 on the Artificial Analysis Intelligence Index, on par with Kimi K3 and up 7 points from GLM-5.2. Once the weights are released it will be tied as the leading open weights model @Zai_org has just launched GLM-5.3, which ties Kimi K3 (60) for the most”
https://x.com/ArtificialAnlys/status/2089830890709135426
Cool MiniMax-H3 use case unlocked for game sprites creation, here’s the workflow: 1. one character design still, on a flat magenta background 2. one 5s H3 clip per move (walk, jump, punch, kick, block…) 3. cut each clip down to its pose extremes 4. key out the magenta, pack”
https://x.com/victormustar/status/2089310616854892818
10 years ago Wang Xingxing was a Shanghai University grad student building a $3,000 quadruped robot for his thesis. Today, the 36 year old founder is China’s first humanoid robotics billionaire. Unitree closed day-2 of trading at 687 yuan, about a $40B market cap. Xingxing’s”
https://x.com/TheHumanoidHub/status/2090516767055757612
That escalated quickly. Unitree began trading on Shanghai’s STAR Market today. Opened at ~1,100 yuan vs the 150.80 yuan IPO price. A ~630% pop that put the market cap at ~444B yuan (~$62B), compared to the ~$9B implied at the IPO price. Trading has since cooled to 883.87 yuan”
https://x.com/TheHumanoidHub/status/2089942820081836513
Europe, two updates for you. Computer History is now available in the EEA, UK, and Switzerland for ChatGPT Pro, Business, and Enterprise users on Mac. ChatGPT can remember activity across your apps and websites, so future conversations feel more personalized and require less”
https://x.com/OpenAIDevs/status/2090487766442512398
The people calling humanoids a bubble just watched Unitree close 460% higher on day one. Market cap at close: ~$50 billion.”
https://x.com/TheHumanoidHub/status/2090088534401220803
GLM-5.3: How Chinese labs keep stride with the frontier
https://www.interconnects.ai/p/glm-53-how-chinese-labs-keep-stride
Pushing the Limits of Serving DeepSeek-V4-Pro – LMSYS Org
https://www.lmsys.org/blog/2026-08-19-deepseek-v4-pro-engine-optimization-h20
I don’t know if this news is real, but it does point out a major issue with “sovereign AI” – if your sovereign models are far behind the frontier, then there is a substantial penalty to using lagging models to compete at frontier applications where the gap is biggest, like cyber”
https://x.com/emollick/status/2090083512653160832
Kimi K3 is now rolled out to over half of the subscription base for included usage, and we’re continuing to expand access today. US and Europe-hosted and zero data retention.”
https://x.com/ollama/status/2090505028998140182
GLM-5.3’s Biggest Upgrade Isn’t the Base Model. It’s the Post-Training Stack. A technical breakdown from Zhihu contributor 欠阿贝尔两块钱 on why GLM-5.3’s gains appear to come largely from post-training scaling… not a new backbone. TL;DR: GLM-5.3 reuses GLM-5.2’s 743B MoE base”
https://x.com/ZhihuFrontier/status/2089977451627847789
GLM-5.3 hits the API at $1.4/$4.4 per million tokens | VentureBeat
https://venturebeat.com/ai/glm-5-3-hits-the-api-at-1-4-4-4-per-million-tokens
A battalion of Booster T2s is getting ready for the World Humanoid Robot Games opening ceremony. Now in its 2nd year, the Games run Aug 22-26 in Beijing at the “Ice Ribbon” Olympic oval. It’s the biggest yet: ~2,056 robots from 660+ teams across 16 countries, competing in 51″
https://x.com/TheHumanoidHub/status/2089761213957947482
Ask a question about any place on Earth and get an answer from satellite data within minutes. That is what Yugen Space is building in Tübingen. I’m featuring Yugen as part of my series around 10 years of @Cyber_Valley, looking at some of the founders and companies growing out”
https://x.com/IlirAliu_/status/2090137218270863411
Galbot has unveiled ET1, its first bipedal humanoid, at WRC 2026. The specs are not revealed, but it looks quite similar to Unitree R1. The robot runs Galbot’s world-action model, built to drive bipedal, wheeled, and heavy-load robots from a single brain. Founded in 2023 by”
https://x.com/TheHumanoidHub/status/2090542321259671604
Galbot is launching its first full-humanoid robot. Until now, the company has only released wheeled humanoids. It’s arguably China’s most relentless embodied-AI company, own models, own hardware and real-world deployments most rivals don’t have yet.”
https://x.com/TheHumanoidHub/status/2089790047126217063
The Honor Lightning humanoid won all three medals at the Beijing half marathon in April. Unitree is engineering its comeback. Founder Wang Xingxing has showcased a new prototype, “Superman” • 28.32 mph top speed, past Usain Bolt’s ~27.8 mph. • 2 meter standing high jump, vs”
https://x.com/TheHumanoidHub/status/2089441104848679281
Most multi-drone systems struggle with one thing: coordination under real-world constraints. A paper in Science Robotics from TU Delft proposes a model-based approach that lets multiple quadrotors jointly move and orient a cable-suspended load. BUT: without relying on learning”
https://x.com/IlirAliu_/status/2089622347829354607
UHS Health Research
https://research.uhs.nhs.uk/news/ai-reveals-hidden-patterns-inside-breast-cancer
DeepSeek is back… and Silicon Valley is terrified – YouTube
https://www.youtube.com/watch?v=xBByvFrqmWU
Leave a Reply