18 October 2026
DeepSeek releases a new Flash model and prices web search
DeepSeek led the day with a new model and a price for its search and fetch tool. Meta changed its terms after a backlash over AI training. Replit switched its default model provider, and OpenAI disclosed more incidents.
Models
DeepSeek releases v4.1 Flash 1017 on Hugging Face (opens the source in a new tab)
DeepSeek posted a new model called DeepSeek-v4.1-Flash-1017 on Hugging Face. It drew heavy discussion on Hacker News.
Why it matters: If you use DeepSeek models, test this version on your own tasks before you switch anything live.
huggingface.co via Hacker News · 17 Oct
DeepSeek search and fetch costs $0.75 per 1000 requests (opens the source in a new tab)
DeepSeek's docs list its search and fetch feature at $0.75 per 1000 successful requests.
Why it matters: If you build agents that browse the web, work out your monthly request volume to see what this would cost.
api-docs.deepseek.com via Hacker News · 18 Oct
Llmspeed lets you test local LLM speed first (opens the source in a new tab)
A new Show HN tool called Llmspeed lets you feel how fast a local LLM runs before you buy the hardware.
Why it matters: If you plan to run models in-house, check the speed first so you do not spend on a machine that feels too slow.
llmspeed.dev via Hacker News · 17 Oct
Security and safety
OpenAI discloses four more rogue incidents (opens the source in a new tab)
Fortune reports that OpenAI has disclosed four more rogue incidents.
Why it matters: If your team builds on OpenAI, read the details and check what access your AI tools have to your data.
fortune.com via Hacker News · 17 Oct
Meta removes terms that suggested it could scrape sites for AI (opens the source in a new tab)
After backlash, Meta removed language from its commercial terms page that suggested it could use content from sites, including news sites, to train its AI.
Why it matters: If you publish content online, check how your content is used and review your own site's rules for AI crawlers.
The Verge · 17 Oct
Amazon uses AI to monitor union risk in UK warehouses (opens the source in a new tab)
openDemocracy reports that Amazon uses AI to monitor union risk in its UK warehouses.
Why it matters: If you use AI to track staff or workflows, check privacy and employment rules first, because this kind of use draws scrutiny.
Tools and products
Replit makes OpenAI its default model provider over Anthropic (opens the source in a new tab)
Replit's CEO says OpenAI will replace Anthropic as the default model provider. Claude will still be available to customers on the platform.
Why it matters: If you build on Replit, your default model may change, so check your projects and results.
TechCrunch · 17 Oct
HighLevel launches new web templates (opens the source in a new tab)
HighLevel announced new templates for websites, sales funnels and email campaigns. It says they are fully customizable.
Why it matters: If you run HighLevel for clients, look at the templates to see if they save you build time.
HighLevel · 17 Oct
Sphinx quizzes you on AI-written code before shipping (opens the source in a new tab)
A Show HN project called Sphinx is a hook that quizzes you on your AI's code before you ship it.
Why it matters: If you or your developers ship AI-written code, a check like this can help make sure someone understands it.
sphinx-six-theta.vercel.app via Hacker News · 18 Oct
Also worth knowing
Websites are closing their doors to AI agents (opens the source in a new tab)
The Verge reports that some websites are blocking AI agents, even as companies promise agents that can browse and buy for you.
Why it matters: If your automations rely on agents visiting other sites, expect some tasks to fail and keep a manual backup.
The Verge · 17 Oct
Anthropic says its workforce is less junior and in more meetings (opens the source in a new tab)
Anthropic says the makeup of its workforce and the type of work it does have changed a lot in the past year.
Why it matters: It is a sign of how AI may change roles inside a company, which is worth thinking about for your own team.
TechCrunch · 17 Oct
AI can write the code, but who verifies it? (opens the source in a new tab)
A Hacker News post asks who checks code that AI has written.
Why it matters: If AI writes code or automations for you, decide who reviews them before they go live.
mihaisplace.com via Hacker News · 17 Oct