AI Tools Recap - January 2025

This is my first monthly update recapping new AI tools, my experiences, and notable industry developments from January 2025.

ChatGPT Operator and o1 Pro Mode: A Mixed Bag

OpenAI launched Operator on January 23 as an autonomous agent preview, prompting me to subscribe to the $200/month ChatGPT Pro plan on January 24. Operator's autonomous web interactions were impressive as a tech demo, but practical usability felt limited, mainly to basic research tasks. I expect OpenAI to make massive progress throughout 2025, and by year's end, we'll likely be blown away by its capabilities.

More impactful was my unlimited access to o1 Pro mode via ChatGPT Pro. This model applies significantly more compute during inference time—likely via parallel runs and/or best-of-N completions—resulting in dramatically improved output quality. I often had 3–5 o1 Pro queries running simultaneously, finding it especially powerful for resolving even the most complex coding challenges that Sonnet 3.5 struggled with.

Repo Prompt: Maximizing o1 Pro's Context Window

In January, I began extensively using Repo Prompt with o1 Pro. This tool allows precise crafting of token-optimized prompts with real-time token counting. By carefully managing my token usage, I could fully leverage o1 Pro's 128K token context window—something that GPT-4o can't practically handle within the ChatGPT interface. I frequently pushed right to the context limit, feeding in substantial code snippets and detailed instructions. This enabled one-shot implementations of full features and complex bug fixes, dramatically improving my development workflow. Even when I had to follow up with o1 Pro to request changes, I never felt like it was struggling with all the context in the chat. It's truly impressive and if you have the right use case, it's worth the money spent. It more than paid for itself.

OpenAI o3-Mini Launch

On January 31, OpenAI released o3-Mini, a smaller and faster reasoning model available in ChatGPT with two modes: o3-mini and o3-mini high. Made immediately available to ChatGPT Plus subscribers, its latency improved by 24% compared to o1-Mini. While not replacing o1 Pro for heavy-duty coding tasks (at least for me), I particularly enjoyed using o3-Mini with Perplexity's reasoning assistant—its speed was impressive even in the high-performance mode, but especially noticeable in standard mode where the reduced latency made a significant difference in rapid iterative work. Via the API - o3 mini is available with the option to use low, medium or high reasoning effort.

Windsurf IDE: Indispensable Development Tool

In January, I upgraded my Windsurf subscription from the early-adopter plan ($10/month) to Pro Ultimate ($60/month) after repeatedly purchasing additional flex credits. Windsurf, an AI-powered IDE built on the open-source VS Code, became indispensable for my development workflow. Its standout feature, Cascade, combines collaborative intelligence with autonomous capabilities, enabling contextually aware, multi-file edits through iterative reasoning.

Windsurf's rapid integration of o3-Mini, DeepSeek R1, and Gemini Flash demonstrated the team's commitment to incorporating cutting-edge models. While I still occasionally turn to Cursor for certain tasks, Windsurf has become my primary development environment. A more comprehensive Windsurf review is on the horizon.

Lindy AI: Client-Proven ROI

I successfully deployed Lindy AI in production for a client in January. This no-code platform for creating AI-powered automations delivered productivity gains that significantly exceeded its monthly cost, solidifying its position in my recommended tool stack. Any AI tool that delivers a multiple ROI through efficiency and time saved is a definite keeper.

Wispr Flow: Enhanced Dictation Efficiency

I began using Wispr Flow (yes, spelled exactly that way) in January, streamlining the drafting of prompts, documentation, and specs. This cross-platform dictation app combines speech-to-text with AI preprocessing and formatting, significantly improving productivity. While I use it on Mac, I've confirmed it works well on Windows as well.

Notion AI: Revisited

I reactivated Notion AI to benchmark its utility against Claude (via Raycast) for refining blog posts. While occasionally useful for retrieving information in my Notion database, the AI model felt limited compared to competitors. More transparency about the underlying model would clarify its capabilities and expected performance, to me, if feels like a small model.

Sunsama for Productivity

January saw me recommit to Sunsama for task management. Its integration with Notion, GitHub, Gmail, and AI-powered time estimation helped immensely with daily planning. The automatic daily highlights feature using LLM summaries provides exceptional retrospective visibility on accomplishments. Expect a deeper dive into my Sunsama workflow soon.

Industry Shifts: DeepSeek R1 & Nvidia Digits

While DeepSeek R1 was notable for its human-like reasoning capabilities at an exceptionally low training cost (~$5.5M), the industry's enthusiastic reaction was all over the internet. The stock market response—including temporary drops for AI hardware providers—seemed particularly misguided given the increasing demand for compute resources from major AI labs.

Nvidia also introduced Project Digits, a $3,000 personal AI supercomputer based on their GB10 Superchip and Grace Blackwell architecture. This compact, Mac Mini-sized device looks like an amazing machine that's probably on every AI enthusiast's wish list this year.

Notable Mentions

Consider subscribing to my email newsletter to stay updated on new posts. I promise to keep emails minimal, only notifying you about new content and significant updates to existing articles like this one.