- What's brewing in AI
- Posts
- š§š¼ o3-mini is around the corner
š§š¼ o3-mini is around the corner
Also: ChatGPT Tasks (7 examples)
Howdy, wizards.
Itās been a busy minute since my last email. I've been working on a new website/tool for you (almost ready so stay tuned), and this week, I started a new role in the marketing team at a tech company.
To balance everything and ensure top quality in this newsletter, I'm going to switch to just 1-2 editions per week. You can expect a Sunday roundup of what's brewing in AI, with an extra update during the week if there's just too much good stuff to hold onto.
PS for following tech & AI on a daily basis, I recommend subscribing to TLDR. They donāt go in-depth, but give you a fresh overview of whatās going on.
Without further ado ā hereās whatās brewing in AI.
DARIOāS PICKS
thank you to the external safety researchers who tested o3-mini.
we have now finalized a version and are beginning the release process; planning to ship in ~a couple of weeks.
also, we heard the feedback: will launch api and chatgpt at the same time!
(it's very good.)
ā Sam Altman (@sama)
8:47 PM ā¢ Jan 17, 2025
Sam Altman just tweeted that theyāve wrapped up the external safety testing for the o3-mini model, and that its release is around the corner.
ICYMI, o3 is OpenAIās new series of reasoning models and is the closest thing to AGI yet. The o3-mini model is, as you might guess, a smaller, faster version of the full o3 model; itās also optimised for coding.
This time, API access it being pushed at the same time as the model hits ChatGPT, meaning builders will be able to start creating apps right from the get go.
ā Why it mattersā ā Exciting times. While benchmarks are flawedāif theyāre at least an indicationā then o3 will probably be able to tackle some complex problems that have been out of reach for AI so far.
PS worth mentioning that while o3-mini has passed the external safety testing, AI alignment remains a topic of huge concern for many of the worldās top experts. Check out this series of recent tweets from an OpenAI researcher who claims they already know claims they already know how to create superintelligence, but donāt know how to control it.
TOGETHER WITH ADS4GPTS
Dario: I just tested ADS4GPTs with my Custom GPT and wrote an in-depth review about it. It's an impressive platform that lets you monetize your AI apps without compromising the user experience. Think of it as Google Ads but made specifically for conversational AI.
Here's the details:
Set up in just 5 minutes with your Custom GPT ā no coding required
Seamlessly insert contextual ads into your AI chat experiences that feel natural and non-intrusive
Integrate also with popular AI frameworks like LangChain, Vercel AI SDK and more
Control the data that leaves your system with a privacy-first approach
Get monthly payouts with performance tracking
DARIOāS PICKS
OpenAI dropped something cool this week: You can now schedule tasks that ChatGPT will do for you and remind you with an alarm. One-time or weekly reminders? Daily news briefing on a specific subject? Daily meal plan? Thereās really a bunch of ways you could use this feature in your daily life.
Hereās some ideas to get your mind going:
Get different types of news digests, e.g. a global news summary every morning, summary of academic papers for your field. With this use case, beware of fake headlines.
Monthly alerts for upcoming concerts of your favourite artists in your area
ChatGPT Tasks is already rolled out to Plus, Pro, and Teams subscribers.
You can configure if you want to receive the tasks by push notification, email, or both (adjust this under Settings ā Notifications).
Thereās also new task view that allows you to manage your tasks on the ChatGPT web version, and you can max have 10 tasks ongoing at any time.

Manage your tasks in ChatGPT by clicking on your profile icon, then Tasks
ā Why it mattersā ā ChatGPT is way smarter than Siri, so I can see myself using this to schedule to-dos and reminders. When they can sort out some of the initial bugs, like the task management page not working properly, I see this becoming a solid feature for the mobile app.
DARIOāS PICKS
Cohere is shifting focus to making useful AI for work; their latest product, North, combines LLMs, enterprise search, and AI agents into a single, secure platform that businesses can deploy, even in private or air-gapped environments.
The deets:
North beats MS Copilot and Google Vertex AI Agent Builder in several accuracy benchmarks
Has a feature called Compass, a highly secure AI search system that quickly locates and indexes text, images, and spreadsheets (in any language)
Makes it easy to deploy AI agents that handle time-consuming tasks across HR, finance, IT, and moreāfor instance, RBC is already exploring a specialized āNorth for Bankingā to meet finance-specific security and privacy needs.
Optimized for strict security and privacy requirements
ā Why it mattersā ā Cohere is shifting their focus from building the smartest model to a hands-on business solutions with AI. Might be a solid move considering the competition in the model layer of AI, and the current massive interest (and spending) in AI-solutions for business, which I think will only increase with the rise of agents that can complete whole tasks.
DARIOāS PICKS

Source: screenshot by testingcatalog.com // DALL-E
Google is working on a feature for NotebookLMāa research assistant that went viralfor its ability to create AI-generated podcasts on the flyāis testing a feature that generates an interactive mindmap from the sources in the notebook. You can click on nodes in the map and NotebookLM will give you a breakdown of the item in the context of the overall topic.
Letās say you ask about AI news, and it shows you the mind map depicted in the screenshot above. If you click on Luma AI (an AI video generator), itās going to break down recent news related to Luma AI.
This feature is still being tested and is in an experimental phase, so it might change before Google eventually does an official rollout.
ā Why it mattersā ā Combining the deep knowledge of AI with visuals like this can make it more accessible and easier to use. Potentially a very useful tool for getting an overview of any topic, with the ability to quickly dive deep in a visual way, without writing a bunch of prompts.
TOGETHER WITH HUBSPOT
Unlock the full potential of your workday with cutting-edge AI strategies and actionable insights, empowering you to achieve unparalleled excellence in the future of work. Download the free guide today!
THATāS ALL FOLKS!
Was this email forwarded to you? Sign up here. Want to get in front of 13,000 AI enthusiasts? Work with me. This newsletter is written & curated by Dario Chincha. |
What's your verdict on today's email? |
Affiliate disclosure: To cover the cost of my email software and the time I spend writing this newsletter, I sometimes link to products and other newsletters. Please assume these are affiliate links. If you choose to subscribe to a newsletter or buy a product through any of my links then THANK YOU ā it will make it possible for me to continue to do this.