Backgroud

When I received my ielts score at 8.19, I still didn’t get my necessary score for speaking part, But the difference between this time and last several times are that I think I already try my best in the ielts speaking part. So, I don’t want to continue to prepare for ietls because it is expensive and filled with random factors.

Meanwhile, My study visa havn’t passed till now and I confirmed I trigger security investigate which will spend more time than regular visa application. I will delay my master program with high possibility if there is no suprise in next week. It is wired it need so much time for ircc to process the visa application, maybe I need other preparations after next week.

As a result, for english, Firstly, I had already ask for a score review for ielts speaking part, I had already submit the application for Cambridge English. In addition, I will change exam from ielts to duolinguo, whose price are more reasonable and are relative easy to pass the exam if I have at least two months to prepare for that.

More important, I am able to start my personal deveop career right now. This week is the first week for my to develop by myself towards to my personal goal even I had quit my job for over three months.

Product

Idea source

When I prepared for ielts speaking test, I asked help for enlish techer online, I found the best services from teacher to me are show the grammar issues of my speaking, actually, I can provide enough long sentence to answer the question because I have already practice these topics for several times. My major problem is is grammer issues, like past tense or present tense, singular or plural, I made so much these kinks of mistakes. So, I want to make a product to point out the mistakes that peole who are learning english made.

I will made it as a discord bot first. Maybe I will build a app for website in the future.

Realize

• receive audio file or real-time speaking audio
• convert audio to text message
• correct the errors in sentences
• display the errors and correction

Process

Right now, it can receive audio messages and output what I want it to display as a discord bot. I am trying to realize the real-time speaking input, which is the area that I am unfamiliar with, consequently, I need more time to learn some new knowledge.

Obtained new information

• receive voice data is sensitive for playforms: there are not offical API for discord.py to receive real-time voice data, I need to ask help from extension, it’s luck that there is this kind of tool.
• Running LLM on only web is possible: it’s not suitable for every LLM, but it is a low-cost way to achieve some specific products. There are also great tools that support it, like web-gpu.
• There are plenty of LLM with small size, which are friendly for personal developer or small business.
• Discord API is so friendly, My love for discord is increasing as I know this platform more and more.
• cloud machine is expensive, I need to spend almost 200 every month if I want to run the product in the future.

Alternative product

Several products are great but there are always some drawbacks

• All of them are expensive
• All of them can not provide best correction display format
• Some of them cannot keep correction function for a long time

day records

• 8.19: organize my though about this product
• 8.20: explore text-to-speech and text correction model
• 8.21:
  • try only js solution and give up.
  • start to connect discord bot
  • ensure text correction model and start to development
• 8.22: realize the whole process though audio segment input
• 8.23: connect real-time speaking and save data, but it isn’t match the whole project. Another solution need.
• 8.24: rest

Reading

Reading can help me to maintain the connection with latest development of techniques, meanwhile, it can provide more fresh information to me.

• levelsio:
  • indie hacker who pursue extreme freedom
  • made numous products but only several of them achieve success, which are enough for him to living and pursue his goal without much life related pressure.
  • most of his successful products are related with AI, which bring lots of opportunities to the internet.
• DeskHub: hardware product of github bars, so fun
• learning english though printing subtitle with youtube
• Run python code in javastript: I believe I will use this tool in the future
• Build products in several platforms though the same code
• What product are we building
• PDF to markdown: It use LLM to realize, which cost so much resources.
• discord music bot: for fun
• real-time face swap video: this is useful in xxx fields, wired.
• write two solutions for leetcode, I want to make some videos if I have extra time.

Thought

• I am happy that I had started my product right now.
• The passion for new product will diminish and flucate with time, it is the most strong period before you start to realize it. It is hard and crucial to maintain the balance between passion and clear thought.
• I should push code to github, it seems like I take part in huge open source project.
• I am too heavy, I need to loss weight
• I start to write english content like this essay, it is nature and comfortable to continuous writing since I started.

Plan for next week

• product:
  • realize the real-time speaking
  • make a landing page
  • publish it in discord channel
• spend some time on exercises
• some time for duolinguo test
• reading and generate new ideas for next product
• record videos for leetcode solutions

2024.8.25