4 November 2026
HighLevel makes GPT-6 Sol the default for Conversation AI agents
Today's news touches tools many small businesses already use. HighLevel changed its default AI model, and there are new privacy warnings about developer tools and Codex chats. Apple and Google also changed how AI works inside apps and schools.
Models
GPT-6 Sol is now the recommended model in Conversation AI (opens the source in a new tab)
HighLevel says GPT-6 Sol is now listed first in the Model menu and is the default for new prompt-based Conversation AI agents. HighLevel says answers are more accurate than on GPT-4.1, at about the same cost per reply.
Why it matters: If you run HighLevel agents, test your key conversations on GPT-6 Sol before you switch, since the accuracy claim comes from HighLevel.
HighLevel · 3 Nov
Qwen 3.8 27B on a laptop can interpret model weights (opens the source in a new tab)
Simon Willison reports that Qwen 3.8 27B, running on his laptop, can interpret model weights.
Why it matters: Capable open models are running on ordinary hardware, which may give you a lower-cost and more private option later.
simonwillison.net via Hacker News · 3 Nov
Post claiming Astra can read and write Qwen weights is called misleading (opens the source in a new tab)
A Hacker News post argues that the claim that Astra can read and write weights to improve Qwen is misleading.
Why it matters: Be careful with big AI claims on social media and check the detail before you plan around them.
honnibal.dev via Hacker News · 3 Nov
Security and safety
Postman accused of auto-installing AI plugins and sharing API data (opens the source in a new tab)
A GitHub report says Postman silently installs AI plugins and tries to share all API data. This is a claim from the report and we have not confirmed it.
Why it matters: If your team uses Postman, check your settings and what API keys or client data could be exposed.
github.com via Hacker News · 3 Nov
OpenAI may be re-enabling model training on Codex chats (opens the source in a new tab)
A Hacker News thread says OpenAI seems to be quietly turning model training back on for Codex chats. This is a user claim, not an OpenAI statement.
Why it matters: If your team pastes client code or data into Codex, check your data settings today.
Hacker News · 3 Nov
Anthropic is working on a way to prove which AI model answered (opens the source in a new tab)
TechCrunch reports Anthropic is building FlashProve. It would let anyone check that an AI response really came from the model it claims to be from.
Why it matters: Proof of origin could help you trust AI outputs in contracts and client work.
TechCrunch
Tools and products
Apple lets Siri AI work across third-party apps (opens the source in a new tab)
Apple says iOS 27.3, rolling out Tuesday, lets developers add Siri AI features to all apps. Siri can use personal context, natural language prompts and on-screen awareness.
Why it matters: If you have an app for clients, voice control through Siri may soon be something customers expect.
The Verge · 3 Nov
OpenAI says ChatGPT search volume is 20 percent of Google's (opens the source in a new tab)
OpenAI says the volume of search queries run through ChatGPT now equals 20 percent of the volume run through Google.
Why it matters: Customers may find you through ChatGPT, so check how your business shows up when you ask it.
The Verge · 3 Nov
Apple Watch beta adds always-on Audio Intelligence (opens the source in a new tab)
The watchOS 27.3 developer beta lets Siri use the watch mic for Siri Recap, which summarises conversations, and Live Rewind, which creates transcripts.
Why it matters: Always-listening features can raise consent questions if you use them around clients.
Enterprise
Heimstaden uses ChatGPT Enterprise and Codex in property management (opens the source in a new tab)
OpenAI says Heimstaden uses ChatGPT Enterprise and Codex to change property management, cut costs and help employees across Europe.
Why it matters: This is a real example of AI in a service business, though the results come from OpenAI.
OpenAI · 3 Nov
OpenAI shows ChatGPT and Codex helping with engineering design (opens the source in a new tab)
OpenAI says ChatGPT and Codex are helping engineers speed up CAD, PCB design, simulation and other physical-world workflows.
Why it matters: If you work with product or hardware clients, AI is moving into their design tools.
OpenAI · 3 Nov
Palantir's US commercial revenue passes government revenue (opens the source in a new tab)
TechCrunch reports Palantir's U.S. commercial revenue was bigger than its U.S. government revenue for the first time.
Why it matters: Private companies are buying AI platforms at scale, so competitors may be moving faster than you think.
TechCrunch · 3 Nov
Also worth knowing
Google will turn off search for schools that reject AI (opens the source in a new tab)
The New York Times reports Google will cut off search for schools that refuse its AI tools. Ars Technica says schools can skip AI but must give up search too.
Why it matters: If you train or coach in education, expect bundled AI to shape what schools can buy.
nytimes.com via Hacker News · 3 Nov
Nexus Mods makes AI disclosure mandatory for new mods (opens the source in a new tab)
Nexus Mods says authors of new mods must declare whether AI has been used in their work.
Why it matters: Disclosure rules are spreading, so check whether your clients or platforms expect you to label AI work.
The Verge · 4 Nov