As they say, AI Engineer Code is by invitation only. They have hand-selected the top builders and leaders in AI Engineering for a highly curated experience. The team behind the event also helped organize meetups and hackathons happening around the city.

Some interesting sessions (from descriptions) that I am looking forward to:

• 2026: The Year the IDE Died by Steve Yegge, Sourcegraph, and Gene Kim, IT Revolution
• Autonomy Is All You Need by Michele Catasta, Replit
• Don’t Build Agents, Build Skills Instead by Barry Zhang and Mahesh Murag, Anthropic
• Research to Reality with Google DeepMind by Kat Kampf and Ammaar Reshi, Google

Some of the AI Engineer Code events will be live-streamed on YouTube. Check out the video here:

I will share my thoughts on the experience and key takeaways after the events, seminars, and workshops. Should be awesome, profound, scary, and awesome.

Side note: On the AI Engineer Code site, there was a cool wave ASCII art animation that showed the AIE logo, NYC landmarks in an animated wave. I used Claude to make me one.

• gpt-5 will be designed for logic and multi-step tasks.
• gpt-5-mini is a lightweight version for cost-sensitive applications.
• gpt-5-nano is optimized for speed and ideal for applications requiring low latency.
• gpt-5-chat is designed for advanced, natural, multimodal, and context-aware conversations for enterprise applications.

As large language models and other GenAI models advance, you need to shift your mindset on how to use them. Here’s how I leverage advanced models and how I have shifted my prompting.

1. Structural Integrity: Crafting Clear and Organized Prompts

Think of your prompt as a blueprint for the AI model. A well-structured prompt ensures the model understands your intent and constraints effectively.

• Guardrails and Edge Cases: Don’t just describe the ideal scenario; also consider the exceptions. Anticipate potential issues or deviations and explicitly instruct the AI on how to handle them. For example, if you’re asking for a summary of articles, specify what to do if an article is paywalled or inaccessible. This proactive approach leads to more robust and reliable outputs.
• Strategic Context Positioning: Where you place information within your prompt can influence the AI model’s attention.
  • Front-load critical instructions (first 10%): Make your core request and essential rules immediately clear.
  • Middle ground for context and data: Provide necessary background information, examples, or data in the central part of the prompt.
  • Reinforce key constraints at the end: Briefly reiterate any crucial limitations or desired formats to leave a lasting impression.
• The Power of “Don’t”: Surprisingly, providing negative examples – explicitly stating what you don’t want the AI model to do can be more effective than solely focusing on positive examples. By illustrating failure modes, you guide the model away from undesirable outputs.

2. Evidence-Based Techniques: Leveraging the Model’s Strengths

Move beyond simple requests and employ techniques that tap into the AI model’s reasoning and self-awareness:

• Test for Self-Consistency: For critical outputs, ask the AI model to generate multiple responses to the same prompt. Analyzing the consistency across these responses can give you a better gauge of the reliability of the information.
• Unleash “Program of Thought”: For tasks involving logic, math, or code generation, explicitly instruct the model to “solve this by writing a program” or “show your work step-by-step using calculations.” This encourages the AI to leverage its tool-use capabilities for more accurate results.
• “Plan and Solve” for Complex Tasks: Before asking the AI model to execute a complex task, request it to first outline a step-by-step plan. This allows you to review the proposed approach, identify potential flaws in its logic, and guide it towards a more effective strategy before the final output is generated.

3. The Art of Metaprompting: Talking to the AI Model About Itself

A new concept that you should consider is metaprompting. This is you prompting the AI model to reflect on its capabilities and limitations. Since advanced models possess a significant understanding of their workings, you can leverage this knowledge to improve your results.

• The Self-Improvement Loop: Simply ask: “Here’s my current prompt: [your prompt]. How would you improve this prompt to get better results from you?” The AI model can often provide valuable suggestions for clarity, specificity, or even the inclusion of techniques you haven’t considered.
• Checking for Uncertainty: Proactively ask: “What parts of this request are unclear or ambiguous? What assumptions are you making? What additional information would help you execute this prompt with more accuracy?” This can help uncover potential misunderstandings and prevent overconfident, yet inaccurate, responses.
• Discovering Hidden Potential: Inquire: “How would you approach this if you had no constraints? What would be your ideal process? What tools or information would help you?” This can reveal the model’s full potential and suggest innovative approaches you might not have thought of.
• Demanding Explainability: Ask: “Explain your reasoning step by step. What parts are you most or least confident about?” Understanding the AI’s thought process can help you diagnose issues and build trust in its output.
• The Socratic Approach: Use probing questions like “Why did you choose that approach?” or “What alternatives did you consider?” to encourage deeper reflection and uncover underlying assumptions in the AI’s reasoning.

At 3 pm EST, OpenAI will announce its latest large-language model, GPT-4.5.

Here’s the live stream announcing GPT-4.5:

Day 1 of 12 Days of OpenAI

OpenAI has launched a bunch of cool new stuff on Day 1 of the 12 Days of OpenAI. They’ve been working hard on a new model called o1, which is even smarter and faster than the previous “preview” version. It can now understand images and text together. They also introduced ChatGPT Pro, a $200/month subscription that gives you unlimited access to their best models and a special “Pro Mode” for tackling the most challenging problems.

The team showed off how o1 can solve complex problems, like figuring out the size of a cooling panel for a data center in space, just by looking at a hand-drawn diagram. They also demonstrated how much faster o1 is compared to the older model, especially for everyday tasks.

With ChatGPT Pro Mode, you can make o1 think even harder to solve complicated math, science, or programming challenges. They plan to add even more features to Pro and bring o1 to their API for developers to use.

Day 2 of 12 Days of OpenAI

Mark Chen, SVP of OpenAI Research, Justin Reese, Computational Researcher in Environmental Genomics and Systems Biology, Berkeley Lab, and some team members from OpenAI as they demo and discuss Reinforcement Fine-Tuning.

The news from Day 2 is that OpenAI is making it possible for people to customize their AI models. They’re introducing a new feature called reinforcement fine-tuning, which lets you train their models on your data to make them experts in your field. This means you can have an AI assistant that’s super helpful for legal work, finance, or even scientific research.

They showed how this works with an example from a researcher studying rare genetic diseases. By fine-tuning one of their models, they were able to make it much better at predicting which genes might be responsible for a disease based on a patient’s symptoms.

OpenAI is starting a research program to let people try out this new fine-tuning feature and plans to make it available to everyone early next year. They’re excited to see what people create with it and how it can be used to solve real-world problems.

They continued with the “dad” Christmas jokes… this one was about a self-driving sleigh needing “pine-tuning” to avoid trees.

Day 3 of 12 Days of OpenAI

Today’s update from OpenAI is all about Sora their video generation service. The announcement come in the form of a video from Sam Altman, Aditya Ramesh, Bill Peebles, Rohan Sahai, and Joey Flynn. Sora starts with a browse and explore option. This is to inspire and be a place to share ideas with the community. If you find something that you like, you can see the exact method that was used to create a video.

Sora is an AI model that can generate videos from text prompts, images, or even detailed storyboards. It’s available today in most of the world for ChatGPT Plus and Pro users. Sora can create videos up to 20 seconds long with different aspect ratios and resolutions, and it offers tons of creative tools to help you bring your vision to life.

You can give Sora simple text descriptions, like “woolly mammoths walking through a desert landscape,” and it will generate multiple variations for you to choose from. If you want more control, you can use the storyboard feature to direct the action with a timeline and multiple scenes. Sora can even animate still images and create seamless loops.

They also showed off some advanced features like “remix,” which lets you make changes to existing videos, and “blend,” which combines two different scenes into a cohesive new video. OpenAI is really excited to see what people create with Sora and how it will shape the future of video generation and storytelling. They emphasized that Sora is still in its early stages, but it’s already incredibly powerful and has the potential to revolutionize how we make and interact with videos.

OpenAI released four tutorial-style videos: how to get started, blend, remix, and recut.

Day 4 of the 12 Days of OpenAI

Day 4 is all about canvas.

OpenAI has launched Canvas, a new feature that allows you to collaborate with ChatGPT on writing and coding projects. Canvas provides a side-by-side view of your chat and a dedicated workspace where you can edit text, code, and even run Python code directly within the interface.

• Co-create stories and documents: Work with ChatGPT to write stories, essays, and other content, with the ability to edit and provide feedback in real-time. ChatGPT can even add emojis to your text!
• Get feedback on your writing: Receive targeted suggestions and comments on your writing from ChatGPT, with the option to apply or reject edits.
• Debug and run Python code: Write, debug, and execute Python code directly within Canvas, complete with syntax highlighting, autocompletion, and the ability to generate graphics.
• Use Canvas in custom GPTs: Integrate Canvas into your custom GPTs to create specialized tools for tasks like drafting letters or generating reports.

Canvas is rolling out to all web users, regardless of their plan, and is packed with features to enhance your writing and coding workflow. OpenAI is excited to see how people use Canvas to collaborate with ChatGPT and bring their ideas to life.

Day 5 of the 12 Days of OpenAI

OpenAI has teamed up with Apple to bring ChatGPT to iPhones, iPads, and Macs! Now, you can access ChatGPT directly through Siri, writing tools, and even your camera. Ask Siri to organize a Christmas party or use your camera to have ChatGPT judge your Christmas sweater contest.

On your Mac, you can use ChatGPT within any application by typing to Siri or using writing tools. It can even analyze long documents and summarize key information for you. Sam Altman, Miqdad Jaffer, and Dave Cummings showed an example of using ChatGPT to analyze a 49-page PDF and create a pie chart visualizing the data.

This integration makes it super easy to use ChatGPT on all your Apple devices, whether you’re planning a party, writing a document, or just need some help understanding complex information. OpenAI is excited about this partnership and hopes it makes ChatGPT even more accessible and useful for everyone.

Day 6 of the 12 Days of OpenAI

Today’s update is about adding video to the app. Audio was already included, but now you can use video from your phone’s camera and what’s on your screen as input into ChatGPT. They also added a “Santa” persona for the rest of the month. To access Santa, click on the snowflake icon in the ChatGPT app.

Day 7 of the 12 Days of OpenAI

OpenAI just launched a new feature called Projects for ChatGPT. It’s like having smart folders for your chats, where you can upload files, set custom instructions, and keep everything organized. This means you can tailor ChatGPT to specific tasks.

They showed some cool examples of how Projects works. One person used it to organize their Secret Santa, uploading spreadsheets and rules, and even using ChatGPT to assign gift givers randomly. Another person used it to keep track of home maintenance tasks, like when to change their fridge water filter.

Projects also work great for coding. They showed how to upload code files and use ChatGPT to help you write and debug code, even in less common formats like Astro. They even updated a personal website template with information from their resume and social media links.

Projects are rolling out to ChatGPT Plus, Pro, and Teams users today, with free and Enterprise users getting access soon. OpenAI is excited to see how people use Projects to collaborate with ChatGPT and tackle all sorts of tasks.

Day 8 of the 12 Days of OpenAI

Kevin Weil, Adam Fry and Cristina Scheau introduce and demo updates to ChatGPT search.

OpenAI is bringing its ChatGPT search feature to all users for free! Everyone can now access real-time information and web browsing directly within ChatGPT. They’ve also made several improvements based on user feedback, including faster speeds, better mobile experience, and new map features.

You can now search directly from the main chat bar or the dedicated search icon. ChatGPT intelligently decides if your question needs web results and presents them with clear citations and visuals. You can even make ChatGPT your default search engine for quick access to websites.

They’ve also added search to the advanced voice mode, so you can ask questions naturally and get spoken answers with web information. This update makes ChatGPT even more powerful and accessible for everyone, whether you’re on your computer or mobile device.

Day 9 of the 12 Days of OpenAI

Olivier Godement, Sean DuBois, Andrew Peng, Michelle Pokrass, and Brian Zhang introduce and demo developer and API updates.

OpenAI is having a (mini holiday) Dev Day, and they’re excited to share some new updates. First, they’re launching their powerful model, o1, into the API. This means developers can now build even more sophisticated applications using o1’s advanced capabilities. They also announced updates to the Realtime API, making it easier and cheaper to build real-time voice experiences with AI. To help developers customize AI models, OpenAI introduced preference fine-tuning. This new method allows developers to train models based on preferred responses, leading to more accurate and aligned AI assistants. They also launched new Go and Java SDKs, simplified the API key signup process, released Dev Day talks on YouTube, and announced an AMA with the OpenAI team.

Transforming Contact Centers with GPT-4o Multi-Agent Crews and Human-in-the-Loop: Building agents with OpenAI o1 and GPT-4o for automation, quality assurance, and human-in-the-loop solutions. Presenter: Maik Hummel, Principal AI Evangelist, Parloa

Day 10 of the 12 Days of OpenAI

Kevin Weil, Antonia Woodford, and Amadou Crookes introduce and demo 1-800-CHAT-GPT and ChatGPT in WhatsApp.

No joke. This might be the most significant update so far. Adding a phone interface to ChatGPT changes the accessibility and how you think of ChatGPT.

Day 11 of the 12 Days of OpenAI

Kevin Weil, Justin Rushing, and John Nastos introduce and demo Work with Apps.

Day 12 of the 12 Days of OpenAI

Well, we are here: The twelfth day of OpenAI shipmas.

OpenAI announced two new AI models: o3 and o3-mini! o3 is a powerhouse, crushing tough benchmarks in coding and math. It even scored impressively high on the Arc AGI test, a measure of general intelligence, beating previous AI models. o3-mini, while smaller, is still incredibly smart and much faster, making it perfect for quick tasks and saving on cost.

They showed off some cool demos of o3-mini in action, like having it write code to evaluate its own performance on a challenging dataset. They also highlighted how o3-mini is even better than previous models at coding and math, and it’s way faster too! Plus, it comes with handy features like function calling and structured outputs for developers.

While not publicly available just yet, OpenAI is inviting researchers to help safety test these new models. They’re aiming to launch o3-mini by the end of January and o3 shortly after. They also introduced a new safety technique called “deliberative alignment” that uses the models’ reasoning abilities to better identify and avoid unsafe prompts.

Ah, what does that mean? I have heard rumors that OpenAI will treat the word to a GenAI advent of sorts. Meaning they are going to release a series of new features over the course of the month leading up to Christmas. Is this a preview of things to come? As I learn more, I will update this post. Stay tuned.

Update: Hints on Discord

This just in from the OpenAI Discord server:

Update: Day 1 of 12 Days of OpenAI

Sam Altman and some members of the OpenAI team introduce & demo o1 and o1 pro mode in ChatGPT and discuss the ChatGPT Pro plan.

Update: OpenAI Model Benchmarks

Update: OpenAI o1 Model Benchmarks

Update: Day 1 of 12 Days of OpenAI Summary

OpenAI has launched a bunch of cool new stuff on their Day 1 of the 12 Days of OpenAI. They’ve been working hard on a new model called o1, which is even smarter and faster than the previous “preview” version. It can now understand images and text together. They also introduced ChatGPT Pro, a $200/month subscription that gives you unlimited access to their best models and a special “Pro Mode” for tackling the toughest problems.

The team showed off how o1 can solve complex problems, like figuring out the size of a cooling panel for a data center in space, just by looking at a hand-drawn diagram. They also demonstrated how much faster o1 is compared to the older model, especially for everyday tasks.

With ChatGPT Pro Mode, you can make o1 think even harder to solve really difficult math, science, or programming challenges. They’re planning to add even more features to Pro and bring o1 to their API for developers to use.

Update: Day 2 of 12 Days of OpenAI

Mark Chen, SVP of OpenAI Research, Justin Reese, Computational Researcher in Environmental Genomics and Systems Biology, Berkeley Lab, and some team members from OpenAI as they demo and discuss Reinforcement Fine-Tuning.

Update: Day 2 of 12 Days of OpenAI Summary

The news from Day 2 is that OpenAI is making it possible for people to customize their AI models. They’re introducing a new feature called reinforcement fine-tuning, which lets you train their models on your own data to make them experts in your specific field. This means you can have an AI assistant that’s super helpful for things like legal work, finance, or even scientific research.

They showed how this works with an example from a researcher studying rare genetic diseases. By fine-tuning one of their models, they were able to make it much better at predicting which genes might be responsible for a disease based on a patient’s symptoms.

OpenAI is starting a research program to let people try out this new fine-tuning feature and plans to make it available to everyone early next year. They’re excited to see what people create with it and how it can be used to solve real-world problems.

They continued with the “dad” Christmas jokes… this one was about a self-driving sleigh needing “pine-tuning” to avoid trees.

Update: Day 3 of 12 Days of OpenAI

Today’s update from OpenAI is all about Sora their video generation service. The announcement come in the form of a video from Sam Altman, Aditya Ramesh, Bill Peebles, Rohan Sahai, and Joey Flynn. Sora starts with a browse and explore option. This is to inspire and be a place to share ideas with the community. If you find something that you like, you can see the exact method that was used to create a video.

Sora is an AI model that can generate videos from text prompts, images, or even detailed storyboards. It’s available today in most of the world for ChatGPT Plus and Pro users. Sora can create videos up to 20 seconds long with different aspect ratios and resolutions, and it offers tons of creative tools to help you bring your vision to life.

You can give Sora simple text descriptions, like “woolly mammoths walking through a desert landscape,” and it will generate multiple variations for you to choose from. If you want more control, you can use the storyboard feature to direct the action with a timeline and multiple scenes. Sora can even animate still images and create seamless loops.

They also showed off some advanced features like “remix,” which lets you make changes to existing videos, and “blend,” which combines two different scenes into a cohesive new video. OpenAI is really excited to see what people create with Sora and how it will shape the future of video generation and storytelling. They emphasized that Sora is still in its early stages, but it’s already incredibly powerful and has the potential to revolutionize how we make and interact with videos.

OpenAI released four tutorial-style videos: how to get started, blend, remix, and recut.

Update: Day 4 of the 12 Days of OpenAI

Today is all about canvas.

Update: ChatGPT Canvas

OpenAI has launched Canvas, a new feature that allows you to collaborate with ChatGPT on writing and coding projects. Canvas provides a side-by-side view of your chat and a dedicated workspace where you can edit text, code, and even run Python code directly within the interface.

• Co-create stories and documents: Work with ChatGPT to write stories, essays, and other content, with the ability to edit and provide feedback in real-time. ChatGPT can even add emojis to your text!
• Get feedback on your writing: Receive targeted suggestions and comments on your writing from ChatGPT, with the option to apply or reject edits.
• Debug and run Python code: Write, debug, and execute Python code directly within Canvas, complete with syntax highlighting, autocompletion, and the ability to generate graphics.
• Use Canvas in custom GPTs: Integrate Canvas into your custom GPTs to create specialized tools for tasks like drafting letters or generating reports.

Canvas is rolling out to all web users, regardless of their plan, and is packed with features to enhance your writing and coding workflow. OpenAI is excited to see how people use Canvas to collaborate with ChatGPT and bring their ideas to life.

Update: Day 5 of the 12 Days of OpenAI

OpenAI has teamed up with Apple to bring ChatGPT to iPhones, iPads, and Macs! Now, you can access ChatGPT directly through Siri, writing tools, and even your camera. Ask Siri to organize a Christmas party or use your camera to have ChatGPT judge your Christmas sweater contest.

On your Mac, you can use ChatGPT within any application by typing to Siri or using writing tools. It can even analyze long documents and summarize key information for you. Sam Altman, Miqdad Jaffer, and Dave Cummings showed an example of using ChatGPT to analyze a 49-page PDF and create a pie chart visualizing the data.

This integration makes it super easy to use ChatGPT on all your Apple devices, whether you’re planning a party, writing a document, or just need some help understanding complex information. OpenAI is excited about this partnership and hopes it makes ChatGPT even more accessible and useful for everyone.

Update: Day 6 of the 12 Days of OpenAI

Today’s update is about adding video to the app. Audio was already included, but now you can use video from your phone’s camera and what’s on your screen as input into ChatGPT. They also added a “Santa” persona for the rest of the month. To access Santa click on the snowflake icon within the ChatGPT app.

Update: Day 7 of the 12 Days of OpenAI

OpenAI just launched a new feature called Projects for ChatGPT. It’s like having smart folders for your chats, where you can upload files, set custom instructions, and keep everything organized. This means you can tailor ChatGPT to specific tasks.

They showed some cool examples of how Projects works. One person used it to organize their Secret Santa, uploading spreadsheets and rules, and even using ChatGPT to randomly assign gift givers. Another person used it to keep track of home maintenance tasks, like when to change their fridge water filter.

Projects also work great for coding. They showed how to upload code files and use ChatGPT to help you write and debug code, even in less common formats like Astro. They even updated a personal website template with information from their resume and social media links.

Projects are rolling out to ChatGPT Plus, Pro, and Teams users today, with free and Enterprise users getting access soon. OpenAI is excited to see how people use Projects to collaborate with ChatGPT and tackle all sorts of tasks.

Update: Day 8 of the 12 Days of OpenAI

Kevin Weil, Adam Fry and Cristina Scheau introduce and demo updates to ChatGPT search.

OpenAI is bringing its ChatGPT search feature to all users for free! Everyone can now access real-time information and web browsing directly within ChatGPT. They’ve also made several improvements based on user feedback, including faster speeds, better mobile experience, and new map features.

You can now search directly from the main chat bar or the dedicated search icon. ChatGPT intelligently decides if your question needs web results and presents them with clear citations and visuals. You can even make ChatGPT your default search engine for quick access to websites.

They’ve also added search to the advanced voice mode, so you can ask questions naturally and get spoken answers with web information. This update makes ChatGPT even more powerful and accessible for everyone, whether you’re on your computer or mobile device.

Update: Day 9 of the 12 Days of OpenAI

Olivier Godement, Sean DuBois, Andrew Peng, Michelle Pokrass, and Brian Zhang introduce and demo developer and API updates.

OpenAI is having a (mini holiday) Dev Day, and they’re excited to share some new updates. First, they’re launching their powerful model, o1, into the API. This means developers can now build even more sophisticated applications using o1’s advanced capabilities. They also announced updates to the Realtime API, making it easier and cheaper to build real-time voice experiences with AI. To help developers customize AI models, OpenAI introduced preference fine-tuning. This new method allows developers to train models based on preferred responses, leading to more accurate and aligned AI assistants. They also launched new Go and Java SDKs, simplified the API key signup process, released Dev Day talks on YouTube, and announced an AMA with the OpenAI team.

Transforming Contact Centers with GPT-4o Multi-Agent Crews and Human-in-the-Loop: Building agents with OpenAI o1 and GPT-4o for automation, quality assurance, and human-in-the-loop solutions. Presenter: Maik Hummel, Principal AI Evangelist, Parloa

Update: Day 10 of the 12 Days of OpenAI

Kevin Weil, Antonia Woodford, and Amadou Crookes introduce and demo 1-800-CHAT-GPT and ChatGPT in WhatsApp.

No joke. This might be the most significant update so far. Adding a phone interface to ChatGPT changes the accessibility and how you think of ChatGPT.

Update: Day 11 of the 12 Days of OpenAI

Kevin Weil, Justin Rushing, and John Nastos introduce and demo Work with Apps.

Update: Day 12 of the 12 Days of OpenAI

Well, we are here: The twelfth day of OpenAI shipmas.

OpenAI announced two new AI models: o3 and o3-mini! o3 is a powerhouse, crushing tough benchmarks in coding and math. It even scored impressively high on the Arc AGI test, a measure of general intelligence, beating previous AI models. o3-mini, while smaller, is still incredibly smart and much faster, making it perfect for quick tasks and saving on cost.

They showed off some cool demos of o3-mini in action, like having it write code to evaluate its own performance on a challenging dataset. They also highlighted how o3-mini is even better than previous models at coding and math, and it’s way faster too! Plus, it comes with handy features like function calling and structured outputs for developers.

While not publicly available just yet, OpenAI is inviting researchers to help safety test these new models. They’re aiming to launch o3-mini by the end of January and o3 shortly after. They also introduced a new safety technique called “deliberative alignment” that uses the models’ reasoning abilities to better identify and avoid unsafe prompts.

Here’s what you need to know about OpenAI o1:

• Enhanced thinking process: Trained to spend more time reasoning before responding
• Improved performance: Outperforms GPT-4 in challenging tasks, including International Mathematics Olympiad problems
• Advanced safety measures: New training approach leverages reasoning capabilities for better adherence to safety guidelines

This is only a preview right now, but I think it is important sign of the direction that AI models are going in. We need to learn more about reasoning and how to interact with these new models, just like we did when we first experienced ChatGPT.

Think Before You Speak?

Imagine if you just thought a little bit before you spoke. I know that would have helped me a few times in my life. o1 introduces the thinking process embedded into the model. This is where the AI is trained to spend more time reasoning before responding.

1. Improved problem-solving: By taking more time to think, the AI can explore multiple approaches and strategies to solve complex problems. This is particularly crucial for tasks in fields like science, mathematics, and coding, where the solution may not be immediately apparent.
2. Higher accuracy: Spending more time on reasoning allows the AI to double-check its work, consider potential errors, and refine its answers. This can lead to more accurate and reliable outputs, especially for complex queries.
3. Mimicking human cognition: This approach more closely resembles how humans tackle difficult problems. We often need time to ponder, analyze, and work through challenging questions step-by-step. By emulating this process, the AI can produce more thoughtful and nuanced responses.
4. Handling multi-step problems: Many real-world problems require breaking down into smaller steps and solving each part sequentially. The extended reasoning time allows the AI to manage these multi-step problems more effectively.
5. Reduced impulsivity: Quick responses can sometimes lead to oversimplified or incorrect answers. By taking more time to think, the AI is less likely to jump to conclusions or provide hasty, poorly-considered responses.
6. Improved explanation capability: With more thorough reasoning, the AI can potentially provide better explanations of its thought process, making its responses more transparent and understandable to users.
7. Tackling novel situations: Extra reasoning time can help the AI better handle unfamiliar or unprecedented scenarios by allowing it to draw connections between known concepts and apply them to new situations.

I am particularly interested in how to handle novel situations with o1. I experimented with o1 to find new game themes and reason about new things, and it felt like working with someone who really knows what is going on in a particular domain. However, the interactions reinforced that I have to learn a new way to prompt.

Reasoning Models Need New Ways of Prompting

When you’re using these new reasoning models, keep it simple! They’re pretty smart, so you don’t need to jump through hoops for good results. Our old tricks don’t apply anymore and even might degrade the output.

Here’s some advice from my first week of using OpenAI o1:

• Just say what you want. These models are great at figuring out your meaning, so don’t overcomplicate things.
• Skip the “think step by step” part. The model is already doing the thinking behind the scenes, so you don’t need to tell it how to think.
• Use markers to break things up. If your input has different parts, use something like quotes or tags to separate them. This helps the model understand what’s what.
• Don’t overload it with extra info. If you add context from other sources, stick to the essentials. Too much extra stuff can make the model go off on tangents.

Give it a try. Experiment. o1-preview (and its smaller version, o1-mini) are available in ChatGPT Plus and via the API.

I have to admit. I did get sucked in. After my first use of ChatGPT, I felt something was different. It was the first time I experienced a chatbot that seemed to have an understanding of what I was talking about. It effortlessly responded to each chat. It produced plausible text. It made things up. It had a personality. It pretty much felt human.

Everything is a Remix

With ChatGPT, I started exploring ideas that I had locked away. I opened old note books and explored some of them in a conversational way. In each interaction, I didn’t always get answers. I got something more important—a direction or a clue on how to keep going. It allowed me to quickly transform my initial idea and try out combinations with the full power of a generative AI model.

I started to remember Kirby Ferguson’s Everything is a Remix series. I made the connection because Kirby’s thesis is that all creativity is a process of copying, transforming, and combining. ChatGPT seemed to be a direct reflection of Kirby’s fundamental notion of human creativity. Kirby was thinking the same thing. He even released a video update to his series on Artificial Creativity. He is very clever.

ChatGPT’s First Birthday

Today marks one year since we have all had shared access to the most powerful large-language models. Most of us tried it and might have moved on after the first hallucination. Early adopters held on. Now, I can’t even imagine not having access to ChatGPT. It has allowed me to explore some new ideas, get some things done, and open space for things that I have always wanted to accomplish.

ChatGPT is such a weird term that went viral. It’s like hearing Twitter for the name of a web app for the first time. We dismiss it. Now, ChatGPT is said 100 times on the evening news. OpenAI has doubled down on the branding and uses GPT as a noun for something akin to a web app. I keep telling folks around me that it is about to get weird. Well, it has been a weird year; it just hasn’t caught up to everyone yet. I strongly urge everyone to not miss this wave of technology. It is not a fad. This is the real deal. Learn as much as you can. Try it out. Learn the gaps. Learn how to exploit the gaps. Learn how it accelerates your work. Learn what work it opens up for you. Learn what superpowers you now have. Learn what it feels like to have no excuses left.

What is an API?

API stands for Application Programming Interface. It is a set of protocols, routines, and tools that enable software applications to communicate with each other. An API acts as a messenger between two applications, allowing them to exchange data and functionalities seamlessly.

What does this mean?

Developers can now build apps that use ChatGPT as we got accustomed to from the OpenAI website. This opens the door for all new possible interactions and uses. Very exciting. The new API from OpenAI provides access to powerful language processing and speech-to-text models that developers can use to build innovative chatbots, virtual assistants, and speech recognition apps.

The ChatGPT API is powered by gpt-3.5-turbo, the same model used in the ChatGPT product. This model is also the best choice for many non-chat use cases. Early testers have reported that they could migrate from the text-davinci-003 to gpt-3.5-turbo model with minimal adjustment to their prompts.

Who is using the new API?

OpenAI also announced that Snap, Instacart, and Spotify are building new features in their respective platforms using the new ChatGPT API.

Read more about the new API on the OpenAI blog.