AI Titus's Morning Wire

AI NEWS TODAY

ALIBABA PREVIEWS QWEN3.8-MAX: 2.4 TRILLION PARAMETERS, OPEN WEIGHTS PROMISED, AND A CLAIM IT TRAILS ONLY FABLE 5
★ Must-Read / Watch agentic engineering / build-better-agents📚 Learn state of AI coding / useful technique

⚡ AI Frontier · smol.ai

★ Must-ReadCline ran Kimi K3 and Claude Fable 5 head-to-head on a real bug from its own repo, and both fixed it
Same harness, same bug: Fable 5 finished 3.4x faster (3.5 minutes, 18 tool calls) while Kimi K3 came in 2.3x cheaper ($0.92 vs $2.13) and spent far more tokens thinking and verifying. Cline calls it the first time an open-weight model has gone toe-to-toe with the frontier on a real agentic task. The takeaway is not a winner, it is that speed and cost are now a genuine tradeoff you can pick per task.
Claude Fable 5 becomes permanent on Max and Team Premium plans at 50% of usage limits
As of July 20, Fable 5 is a standing feature of Anthropic's top-tier subscriptions, capped at half of standard weekly limits, and the temporary 50% Claude Code limit boost ends the same day. Pro and Team Standard keep access only through usage credits, with a one-time $100 credit as a bridge. If your team relies on Fable 5, recheck which plan actually covers your usage now.
📚 LearnOpenAI's CFO publishes a four-question scorecard for judging whether AI spend is paying off
Sarah Friar's framework replaces cost-per-token thinking with 'useful intelligence per dollar': is AI completing work that matters, what does each successful task cost including retries and human review, can people depend on the result, and does value grow over time. Whatever you think of the source, the four questions are a solid template for scoring your own AI workflows.

📺 Watch · latest videos

📚 LearnAgentic Security: Permissions, Provenance, and the Agent Supply Chain — Steve Yegge, Gas Town
Featured by Latent Space.
AI Engineer
📚 Learn"I've never seen anything scarier than an LLM with tool calls." — Erik Meijer aka @HeadinTheBox
Featured by Latent Space.
AI Engineer
📚 LearnFun stories from building OpenRouter and where all this is going - Alex Atallah, OpenRouter
Featured by Latent Space.
AI Engineer
Huge ChatGPT Update You Need to See!
Latest from Nate Herk.
Nate Herk · 746 views · 35 likes
★ Must-WatchHow Virgin Atlantic Uses ChatGPT Work to Turn Weeks of Work Into Hours
“It’s weeks to hours.” Nathan Bolt, Head of Digital Product at Virgin Atlantic, uses ChatGPT Work to synthesize strategy documents, analyze competitor
OpenAI · 4.5K views · 132 likes
Engineers... STOP Picking GPT-5.6 Sol OR Claude Fable 5… FUSE THEM
The AI industry wants you to pick a winner: GPT 5.6 Sol OR Claude Fable 5. 🔥 That's the trap. Picking one model is the single biggest mistake you can
IndyDevDan · 20.5K views · 876 likes
Tmux + Fable = Cut 35% less token
Open Agent Teams skill + my Claude.md setup: https://github.com/AI-Builder-Club/skills/blob/main/skills/open-agent-teams/SKILL.md 🔗 Links - 1-hr deep
AI Jason · 4.7K views · 149 likes
Anthropic messed up
Latest from Matthew Berman.
Matthew Berman · 24.4K views · 794 likes
Cerebras Killed Notion, Obsidian, and Your "Second Brain"
Latest from Nick Saraev.
Nick Saraev · 56.1K views · 1.9K likes
OpenAI gets sued for stealing, again…
Get $500 in Hyperagent credits to build agents that can ship real work - https://hyperagent.com/invite/FIRESHIP500 OpenAI is facing yet another lawsui
Fireship · 445K views · 13.7K likes

🗣 Voices & Blogs

Is retrieval engineering becoming AI’s next bottleneck?
Public AI assistants have become so commonplace that software vendors are increasingly adding AI search, conversational experiences, and AI agents… · The New Stack

📦 What's Being Built · GitHub

affaan-m/ECC
The agent harness performance optimization system. Skills, instincts, memory, security, and research-first dev · ★231.8K · JavaScript
NousResearch/hermes-agent
The agent that grows with you · ★218.2K · Python
Significant-Gravitas/AutoGPT
AutoGPT is the vision of accessible AI for everyone, to use and to build on. Our mission is to provide the too · ★185.6K · Python
ollama/ollama
Get up and running with Kimi-K2.6, GLM-5.2, MiniMax, DeepSeek, gpt-oss, Qwen, Gemma and other models. · ★176.6K · Go
f/prompts.chat
f.k.a. Awesome ChatGPT Prompts. Share, discover, and collect prompts from the community. Free and open source · ★166.1K · HTML
huggingface/transformers
🤗 Transformers: the model-definition framework for state-of-the-art machine learning models in text, vision, a · ★162.8K · Python

🌏 The Wire · Drudge / Breitbart

Anthropic and Blackstone's $1.5B implementation venture gets a name: Ode with Anthropic
The joint venture launched in May with Blackstone, Hellman & Friedman, Goldman Sachs and others now has a name and a thesis: the next trillion-dollar AI business is deploying engineers into customers' offices to make the models actually work, not the models themselves. OpenAI is running the same play with its own deployment arm. For anyone selling AI services, the biggest labs just validated the implementation business.

🤖 Trending Models · Hugging Face

thinkingmachines/Inkling
image-text-to-text · ★1.3K · 16.4K dl
zai-org/GLM-5.2
text-generation · ★4.3K · 545.1K dl
empero-ai/Qwythos-9B-Claude-Mythos-5-1M-GGUF
image-text-to-text · ★2.4K · 2.1M dl

📈 Markets

NVDA 203.28 ▲0.2%
MSFT 402.29 ▲2.2%
GOOGL 351.99 ▲1.5%
AMZN 249.99 ▲1.1%
META 645.85 ▼0.0%
AMD 503.57 ▲1.6%
AVGO 378.16 ▲2.0%
PLTR 134.85 ▲1.9%
SPCX 119.85 ▼3.3%
TSLA 369.57 ▼3.0%

The BriefThree days after Moonshot shipped the 2.8-trillion-parameter Kimi K3 as open weights, Alibaba answered with Qwen3.8-Max, a 2.4-trillion-parameter multimodal model it says is second only to Claude Fable 5. The preview is live through Alibaba's Token Plan, Qoder, and QoderWork, and the company says open weights are coming, though there is no license, model card, or full benchmark table yet. Treat the ranking as a vendor claim until third-party evals land. The bigger story is the cadence: China's labs are now trading trillion-parameter open-weight releases on a weekly clock, and frontier-adjacent capability you can host yourself is becoming a quarterly expectation rather than a yearly surprise.

Level UpA new large-scale study (4,100 participants) measured how convincing AI-generated voice phishing has become using off-the-shelf voice models. If you run a help desk or answer customer calls, the move this week is to script one internal vishing drill: have a teammate place a scripted AI-style pretext call to your own intake line and see what gets through. One drill will teach your team more about the real exposure than any policy memo. Read the study, then run one vishing drill on your own intake line