How I Saved Over $400 a Year by Investing Just $20 in DIY Subscription Alternatives

From social media personalities to prominent business leaders, there’s a growing narrative that AI might signal the decline of Software as a Service (SaaS). While a reduction in subscriptions sounds appealing, the reality I observe suggests otherwise; the number of software subscriptions appears to be increasing yearly, often at inflated costs.

Determined to explore alternatives, I set out to see if I could use AI to substitute some of my existing subscriptions. I ventured into utilizing my $20 Claude subscription to generate coding alternatives for various tools I rely on daily. Surprisingly, this experiment yielded a savings of over $400 annually.

Below are the subscriptions I managed to replace using Claude.

I Saved $144 Annually on Grammarly

Leveraging Clever Prompts and Claude’s Features

A significant discussion point regarding AI revolves around its capacity for content creation. Personally, I’m not fond of AI-generated texts. They often come across as mechanical and insincere. However, LLMs (large language models) do possess a strong command of grammar. While there have been instances of nonsensical outputs, serious grammatical errors are rare.

Consequently, Grammarly was the first subscription I eliminated after experimenting with LLMs. These models can easily detect most grammar-related issues, including typos and punctuation errors. Additionally, LLMs can understand context, enabling them to provide more relevant corrections. For instance, if I accidentally typed “can” instead of “can’t” in a contextually opposite sentence, an LLM would catch this subtle discrepancy more effectively than Grammarly.

The primary challenge with using an LLM instead of Grammarly lies in usability. Copying text into a chatbot for corrections is an inefficient process. Grammarly allows users to see each error and decide whether to accept suggested changes. However, I found a workaround by using Claude Artifacts. This allows me to submit my writing to the LLM and request an Artifact that highlights the grammatical mistakes and provides corrections alongside them. Other platforms like ChatGPT Canvas can be used similarly.

I Saved Another $144 Annually on Wispr Flow

Combining a Python Library with an Open-Source Dictation Model

As someone who writes professionally, I don’t particularly enjoy typing. Over time, I’ve developed wrist pain that makes lengthy typing sessions uncomfortable. Additionally, my typing speed is significantly slower compared to my speaking speed. This has led me to prefer dictation tools.

I’ve used dictation software long before the AI boom, with Dragon NaturallySpeaking being a primary choice. Nowadays, numerous open-source models are available, like OpenAI’s Whisper and NVIDIA’s Parakeet, along with various subscription-based tools offering user-friendly interfaces.

Wispr Flow has emerged as one of the top speech-to-text applications, which I appreciated. However, I grew frustrated with its subscription-based model, especially since it utilized open-source technology. I realized I could create a similar application using these tools and save money.

I enlisted Claude to design a solution, which included RealtimeSTT—a real-time transcription tool—and faster-whisper, an enhanced Whisper implementation. Using Claude, I also developed a PowerShell script that seamlessly integrates these resources and starts automatically when my system boots up.

This script features both push-to-talk and live transcription modes. By pressing F9, I can record my voice, which will transcribe my speech and save both audio and text files. Alternatively, using the live mode allows for transcriptions to appear in real-time as I speak.

I Saved Another $120 Annually on Rize

Creating an HTML Dashboard and Integrating It with ActivityWatch

I discovered Rize while participating in an online productivity course. It’s a time-tracking application for Windows and macOS that logs user activity in the background. Its AI component categorizes activities into work, leisure, and personal tasks, providing insights about productivity.

While I found it valuable, the $120 annual fee felt excessive for what it offered. Additionally, I was uncomfortable with sharing my personal data with a third-party service. Instead, I opted for ActivityWatch, a more private, open-source time tracker.

Though ActivityWatch lacks AI features, I addressed this by processing its raw data with an LLM, which can categorize and tag my activities according to my preferences. This analysis feeds into a customized HTML dashboard that I developed with Claude’s assistance.

An Affordable AI Subscription Can Yield Significant Savings

While it may seem like I’m substituting a few SaaS subscriptions with yet another AI service, that’s not entirely accurate. For just $20, you can build the projects I mentioned without incurring any further costs. Once created, any apps or scripts developed with AI become your property to use at will. This underscores that AI’s true potential lies in enabling users to create offline solutions rather than solely relying on them as conversational tools.