In partnership with
☀️ TRENDING AI NEWS 🤖 Drex Max v1: Nace AI open-sourced a 27B multimodal decision model that scores 67.52 and runs locally on Apple Silicon. 🛠️ dsh-superagent: DeepSeek's MIT-licensed router-worker coding agent cut simulated p95 task time from 29.5 to 15.7 minutes. 🚨 Anthropic Credit: Blackstone expects a high 'junk' rating for Anthropic after its IPO, per an FT report. 🏢 Market Wobble: A tech sell-off knocked nearly $1 trillion off US markets on Friday after weak earnings. |
Two open-source releases landed over the weekend that attack the same problem from opposite ends: one shrinks serious decision-making onto a Mac, the other makes coding agents work in parallel instead of in line.
| 🛠️ Tool of the day | |
Julius AI - an AI workspace that analyzes your data and builds charts, dashboards and presentations from plain-language questions. It helps analysts and busy professionals who live in spreadsheets. It has a free plan.
| ⚡ Quick hits | |
Google, Microsoft, OpenAI and Anthropic will honour one shared robots.txt opt-out for AI training. Read more →
Zest launched an AI agent that finds restaurants and books reservations over text. Read more →
Ollie, a privacy-focused AI assistant, is shutting down. Read more →
| 🤓 AI Trivia | |
Which Google DeepMind model did the company's first orbiting TPU satellite use to write a haiku about being in space?
The answer is hiding near the bottom of today's newsletter... keep scrolling. 👇
| 🤖 Nace AI Puts a 27B Decision Model on Your Mac | |
Open weights, local inference, real benchmarks
Nace AI open-sourced Drex Max v1, a 27B-parameter multimodal decision model that scores 67.52 and is built to run locally on Apple Silicon. That makes it a notable entry in the growing open-source AI pile, where the pitch is no API bill and no data leaving your machine.
A decision model is aimed at picking among options rather than just chatting, which suits workflows like triage, routing and planning. Running it on a laptop-class chip means teams with strict privacy needs can experiment without standing up a GPU cluster.
The bottom line
If you have been waiting for a capable multimodal model you can run entirely on your own Mac, this is worth a weekend of testing, though the 67.52 score deserves your own evaluation on your own tasks.
Why are so many brands shifting ad spend to CTV?
People spend more time streaming than they do with cable, social media or video games, and advertisers know it. In fact, according to Advertiser Perceptions and Premion's 2026 CTV/OTT Advertiser Survey, 70% of advertisers are raising CTV budgets in 2026, with much of that budget diverted from social advertising.
For marketers who want a presence in streaming, Roku is the obvious place to go, with scale that speaks for itself. With more than 100 million households representing 47% of all streaming time in the US, Roku easily has more reach than every competitor.
With Roku Ads Manager, you can reach viewers across more than 100 premium apps in a single buy. Combine your customer data with Roku's insights to reach only the households you want, then measure and optimize to your heart's content.
Want to give it a try? Unleash your first campaign in just four steps: (1) Create your account, (2) define your objective and target audience, (3) upload creative, and (4) launch.
First-time spenders get a $5,000 match on the first $5,000 spent with code BEEHIIV5K. Terms apply.
| 🛠️ DeepSeek's dsh-superagent Splits Work Across Parallel Workers | |
A router that spawns worker sessions
DeepSeek released dsh-superagent, an MIT-licensed research prototype that turns GitHub and Feishu requests into parallel worker sessions. A router reads the incoming request and hands pieces to workers instead of processing everything in a single queue.
In DeepSeek's own simulation, p95 task completion fell from 29.5 minutes to 15.7 minutes. That is a vendor simulation, not an independent benchmark, but it points at where coding agents are heading: orchestration and concurrency matter as much as the underlying model.
The bottom line
Treat the 15.7-minute figure as a hypothesis rather than a result, but the router-worker pattern is easy to copy and cheap to try if your team is drowning in tickets.
| 🏢 Anthropic's Post-IPO Debt Could Be Rated 'Junk' | |
A forecast, not a verdict
Following the IPO filing we covered over the weekend, the FT reports that credit agencies are expected to give Anthropic a high junk rating. The forecast comes from Blackstone, so it is a prediction rather than an official rating, but it matters because a junk rating would make it costlier to raise the money that funds the company's massive spending.
The mood around financing is already tense. Friday's tech sell-off knocked nearly $1 trillion off US markets as investors questioned whether billions poured into AI can generate returns quickly enough. TechCrunch also reports that OpenAI has reportedly lowered the funding target for its new round, and that Nvidia is weighing cosigning a loan for Nscale's West Virginia data center.
The bottom line
The labs' costs are rising faster than investors' patience, so expect more creative financing and more headlines about who is propping up whom.
| 🍎 Apple's CEO: No 'AI for AI's Sake' | |
Apple's new CEO defended the company's current position in the AI landscape, saying Apple does not want to ship AI for AI's sake. It is a pointed contrast to the spending race, and it lands as Apple prepares to bring Apple Intelligence to China with iOS 27.3, with a beta due November 4 and a full release in December.
One caveat from The Verge: the China launch does not include the new Siri AI, only a subset of Apple Intelligence features.
| ⚠️ Courts and Copyright Heat Up | |
A judge allowed an NAACP lawsuit accusing the SpaceX-owned xAI of violating the Clean Air Act over data center pollution to move forward. Meanwhile in Australia, experts warn that letting AI companies scrape copyrighted material could breach the constitution and trigger a huge compensation case, and artists in the prime minister's own electorate are pushing back.
Need a website while you read all this? Our sponsor 60sec.site is an AI website builder that gets you online in about a minute. And for more AI news every day, visit dailyinference.com.
| 🌎 Trivia Reveal | |
The answer is Gemma 4! Google says it booted up the satellite's TPUs and the on-board Gemma 4 model followed a prompt to write a haiku: 'Blue marble below / Data streams in silent grace / Starlight feeds my core.'
| 💬 Quick Question | |
Would you rather run a capable AI model locally on your own machine, or keep using cloud APIs for the convenience? Hit reply and tell me why - I read every response!
That's all for today - see you tomorrow!
| 🎁 Share Daily Inference | |
Share your link with friends. Get the AI Tools Starter Kit at 1 referral and the AI Insider Briefing at 5.