Monday, June 19, 2023

CovGPT Progress 2

总算把DB的update部分弄好了。

这个部分耗时挺长 (前后4天),主要没经验,

1. 需要form prompt

2. 再来GPT 一般work, 就是data formating 要注意

再来是有几个Big bugs


  第19天

7:14am waketup

7:21am finish yesterday

7:52 jmg + breakfast

8:10 breakfast

8:21 this blog

9:26 xf

9:35 design

10:02 fix the first item select issue (the execution sequence due to async and 

10:46 investigate the big bug3 and draft and stackoverflow question and ask

11:08 put a ondelete screen

11:29 lunch

11:59 meeting

12:11 news

13:47 back

14:21 shixiong+mother+transfer money+KW reply

14:42 write code the extract content

15:05 popup delete code done

15:12 delete logic done

15:43 delete option and with right option selected

15:49 question container hover

16:06 document on ts and to start on the question

17:40 add question done

  chatgpt is not getting the best suggestion - you need to ask it to put in onSubmit of form, the more specific the better. If you are going to ask where is the best place to put the code, as it first, before you form instruction. For GPT, focus is good, smaller amount of clearer task will increase the ROI

17:49 make it read from json

18:00 update for incomplete portion

18:12 dinner

18:55 Baby time

19:06 get questions done

19:49 when empty string and adjust center column

21:28 complete delete questions and delete scenarios named rectification

  第18天

6:40pm wakup

6:56pm see news (7 mins) tidy loom (3 mins) see loom competitive (7mins)

7:25 jmg

7:49 till verify token

8:19 bath

9:30 go baby

10:45 back from kid's father's day

15:20 fix the bug bugs  (first: mine, second: gpt4)

17:19 wakeup

17:30 start

17:50 getCategory

18:00 getAllcategories done

19:11 Dinner + submit success closing 

19:30 figure out what is the best place to put error and ask Ilya, as well question subbmitting

21:15 exercise

21:54 display error and reload scenario

22:12 refreshing listing

Pretty slow- big bug

没事,总有学习点高的地方








  第17天

感觉周天时间多点

7:02am wakeup

7:32am jmg

8:25am back

8:51am organized though

9:12am hair cut and bath

11:33am finished all admin stuff and pd

6:00pm back

5:40-6:00 dinner

6:00-6:23 write KW

Ahem...too much emotional turbulent

The right way:

1. Train your listening skill (how this can help)

2. Handle this in a profesional way

3. When you don't get the data

               a. see what are the thing you can help (time-bound, plan for the worst)

               b. assume everything will come out best (it's the case normally, hope for the best)

               c. don't waste too much time on it.

          4. You need to manage your emotion, don't let your emotion waste time for nothing.

7:24 starting

7:51 no starting

7:51-9:00 finish UOB, learning mfcc + spectogram\

9:16 start react

9:45 change popup windows to the right one

10:11 finish all services function


第16天

5:22am yesterday sleep outside 11:18pm

5:59am finish all deletion (not familiar with deletion, and update create scenario)

6:06am finish authorization

6:28am download yitang and sub course

8:35am wakeup

--reflection-- prioritize 生活, prioritize修行, make this a long term activity

more important is 生活smoothness

9:30 go

10:50 back for awhile, half way list scenario

1:30 back

1:39 get scenarios and questions done

2:06 wash and start listen lesson

7:30 back

8:59 sleep, shixiong course, tbsn letter, see some video, blah..



第15天

6:02> wakeup and 金母功

6:50 write blog

6:59-7:19 integrate with openai and abstract it into a class

8:00 Icon with tooltip

8:22 add close button (fix the absolute position and also the z-index)

8:35 disbled send button accordingly

8:38 breakfast

8:46 done

8:52 add active font color

9:30 half way through category deletion (make it a rule, when you can code in one minute, dont rely on chatgpt, otherwise, you will be a slave of it!)

10:27 back from william parent's day (key, make him self-reliance, and improve on communication)

10:39 delete option worked

10:46 change add new scenario button

3:15pm back

3:45pm bath etc.

4:54pm sleep

5:22pm see email and news

5:57pm ask question 

6:30pm still thinking about it

6:30pm-7:40pm practise and after than eating

7:45pm moving the relevant folder to covgpt

10:14pm get the initial create delete done in db

10:39 get all the logic, remain

deleteScenario

deleteCategory

with authenticaiton to be doleted






第14天

6:45-6:56 Have stackoverflow question wrote and have GPT answered, ultimately, I got two classes QuestionInput and QuestionInputWithVoice merged together and solved problem. Self-solved? Might take 3-4 hours.

6:58- 8:45 exercise

8:45-9:15 find all volo notes

9:15-9:48   cut hair+bath

-11:00 mantra

11:00-12:37 

#send the ai answer
#ask 物业
#ask shixiong
#cancel slack
meeting

12:37-1:34 wakup

1:46 merge the voicebutton into questioninputwithvoice

2:08 tidy desktop and try out the button volume

2:26 initial volume console log is done

2:36 some finetuning on the volume

2:46 volume with audio visual finish initial

3:20 still deal with volume smoothing problem

4:54 back from massage

6:22 finish handling the volume smoothing problem and extract the svg class out and fix the ide issue

7:12 discuss with ilya, having dinner

7:13-7:41 finish the button centralized

7:49-9:34 everything for sending finish, spend a lot of time to abstract questioninputwithvoice, not worth it.

9:34-10:21 server side connected (gpt teach how to get query)





第13天

TTS day

7:07am-8:19am investigate stt, breakfast, bath, share, to chant

9:30am until

11:03am start middle wait for the computer recovery see video

2:32pm record loom video for meeting, nap, chat with  Agung, think of NJ, plan for stand up, think of deepgram

to 4:19pm finish the UOB task

7:30pm-10:20pm finish the STT (using windows component (just 5 mins with GPT!))


Next
#Plan for Watson NLP
#Plan for Standup


第12天

6:56am-7:20am

why background change

why the cloumn widdth/height not right

7:50-7:55am 

give stackoverflow question with help of gpt

eat, bath, chat

9:35am - 12:42pm

the stackoverflow fixed problem

remaing autheticatoin page issue (gpt helps a bit)

1:15pm wakeup-3:06pm finish the get serversideprops redirection (gpt helps alot, this is easily a 1-2 days - task, since a lot of learning is required - watch video etc.)


3:06pm-5:07pm make redux toolkit provision with meetings (not finish yet, but not worth it to continue at this moment)

5:07-7:17pm    dinner and finish all authentication issue (oauth, delete extra pages, router.replace) (gpt not helping much here


summary (help a lot)

1. stackoverflow help

2. serversideprops learning and implement



第11天

下面一小时+学习试错, 平常6-10小时

- fast api testing

- mongodb.js done


till 2:09pm (with all meetings)

- sign in, sign up, forget, reset password done


3:57pm-5:14pm fix all deployment issues

5:14pm - 6:20pm fix authetication flow issue (with dinner)

8:30pm-10:15pm fix remaining

1. after sign up not login

3. google sign in issue

4. google sign up issue


感悟:

1. 现在加学习试错都挺快了,以后没有这个部分就更快了

2. 发现了这是个accelerate process, 本来快,后面还会更快

3. 现在都是文字学习,以前一个command 没搞懂,需要各种找 - 搜寻,看video, 问stackoverflow, 自己试,现在都不需要了,一个问题就搞定。而且看文字是最快的学习方法。马不停蹄的学习开发。

4.开发整体来说是加快6倍到10倍,感觉后面还会继续放大 - 因为有了utilty, 更加熟悉chatgpt(怎么问问题), 和language 本身(减少学习和试错时间)。

5. 傍晚时间会给自己一个rest (驾车,运动 等), 感觉这样很好调和比较紧的工作,和生活。



5 days learning

 2 June - 7 June  Understand the entire architeture and do the first dropdownbox

5 days coding (all frontend)

8 June 9 commits

9 June 3 commit (writing python code)

10 June 8 commits

11 June 10 commits

12 June 14 commits


感觉越弄越快了

没有chatgpt, easily 2 months, learning 1 month, UI 1 month

UI 部分还有前面的code可以参考,不然要花更久。

现在花了10天。


后面就是 backend 了







了解和不了解

我在决定是否需要提升自己时,如果我只是从自己已知的信息出发,而对于职业晋升还没有完全了解,为什么需要思考领导者的角色呢?当你对职业有了更深入的了解后,你就不再需要提出这样的问题了,因此你提问也没有什么用处,因为你尚未获得完全的了解。这意味着你所采取的方法和实施方式也未必是正确的。只有在你获得充分了解之后,你才能够做出正确的决策,此时再去提问就是合适的。



Original

当我如果是从认识的方面来知道确定这题需不需要promote自己,但是还没有完全的了解这个职业上升,为什么需要though leader的,那当你了解之后,你就不需要问了,所以你问了也没有用,因为你不了解,所以你做的方式,实施的方式也不一定是对的,当你了解之后,你不问的时候,那时候才是对的了。


构建飞轮

在路边上清理垃圾这个问题上,为什么如果我不逐点捡拾的话,肯定是无法将其清理完全?而为什么路边却如此清洁?这是因为它存在一个"飞轮效应"。路边的清洁不仅仅是商业上的好处,它会促使政府愿意投入资金雇佣人员来进行清理。而清洁的环境会进一步吸引更多人投资,这是"飞轮效应"中的一个产物。一旦你建立了这个"飞轮效应",其他人也会愿意将资金投入其中,因此你必须创造这个"飞轮效应"。


Original
关于路边上面的打扫的垃圾,为什么我如果没一点一点自己去捡的话,肯定是捡不完的,但是为什么那么旁边那么清洁,是因为它有个flying wheel,路边的清洁不是就是商业的好,会导致政府愿意花钱去请路边的人清洁,而清洁会导致这更多人愿意来投资,它是一个flying wheel里面的其中一个产物,你有了flying wheel之后,其他人就会愿意把这个钱注入在里面,所以你必须要制造的是flying wheel。


进步六倍

 实现六倍进步,是你的目标,这种提升生产力的点,能够轻易吸引他人使用你的产品,因为你能够帮助他们实现六倍的进步,而不仅仅依靠不断加倍的努力。

因此,我常常思考如何让ChatGPT进一步实现六倍的增长。通常,仅仅加倍加倍会使你的生活更加紧张,所以你不如考虑如何采用其他方法实现六倍的增长,而不是将自己的任务同时绷得更紧。你应该思考如何利用像ChatApp这样的工具来实现六倍的增长,这是一种能够帮助你实现这一目标的方法,而不是将任务并行进行。

达到六倍的进步,这是你应该思考的,这样不仅能够吸引他人使用你的产品,还能实现真正的生产力提升,而不仅仅是不断加倍努力所能达到的。

original

进步六倍 还有就是我常常想着chatgpt要怎么样再继续加倍而我想的一般都是加两倍加两倍的话基本上会把你的生活绷得更紧所以你还不如情愿放两倍你应该想的是用什么方法给他加个六倍而这个六倍不是靠绷紧能够完成的就好像需要去chatapp这种tool才能帮你就是实现呃完成六倍儿你要想的就是完成六倍而不就是而不是parallizing your job, 两件事同时进行,把自己蹦的更紧的事物上面。 进步六倍,是你要达到的,这个productivity gain点,这样子的话,而且很容易吸引别人来用你的东西,因为你能帮他们进步六倍,而不是单靠这种effort double effort这种。


分阶段练习 - 他练他的,我练我的

 在修行过程中,我们需要逐步前进,首先修炼自己,摒除周围的干扰,专注闭关修行,培养专注力,然后再修炼利他菩提心。口语练习亦然,我们需要一步一步,分阶段地训练不同的部分,首先是……

首先,倾听并清楚理解他人的表达和意图。 其次,清晰明了地表达自己的观点。 第三,既能主观表达又能说服他人,同时通过说服他人来塑造自己的领导者形象。 第四,这是应对异议的能力,因为似乎应对异议还需要内心的坚韧训练。

当你将前面部分的能力发展到很高水平后,对于坚韧的需求自然会减少,这是一种智慧的修行方法。

经常问自己如何更快速地提高表达能力,我认为也是一步一步地练,没有连接好的一二,练习三就会很困难。 虽然我们的目标都是零,但是你的一二三四这些因素会让你难以达到零,所以你需要逐步练习,这样会更快地达到目标。 有些非凡的人,他们在前世已经将这些部分练好了,因此能够直接达到零。而我们不能从他们身上学习,因为我们其他部分还没有真正训练好,就好比一个天生的奥林匹克选手和一个后天连基本运动都练不好的人,后者需要去练习奥林匹克选手,完全是不同级别的存在。 他悉心修行,我专注自我。

Original

我们修行的话,就是需要一步一步来,先修自己,先把周围环境去掉,然后专心的闭关,先修专一,然后再修,利他菩提心,那练口语一样,我们要一步一步啊,分阶段的把不同的部分练好,第一个部分就是。

这个 第一部分就是听听清楚别人表达什么,别人的话下面想要表达的意思。 第二点清楚的表达, 第三点,你能够主观,但是又能说服别人,同时说服别人的表达就是有自己的意见,他们教会加你thought leader。 第四点才是这个objection handling,因为好像objection handling还需要你自己心里面的这个resilient的训练。 等你把前面部分的能力发挥到很高之后,那个resilient的需要自然会减少,所以这是一个很有智慧的这个练法。 这样的我常常问自己要怎么样更快速的把这个出事方面练好我觉得也是一步一步练,一二没连好,练3难度就会很高。 虽然我们的这个目标都是零但是你的一二三四这些东西会让你很难迈向0,所以你要一部分一部分的练,这样会更快速到。 有些比较厉害的人,他是前世就已经把这些部分练好了,所以他直接能够练0,而我们不能跟他们学,因为我们的其他部分都还没有真正的练好,就好像一个天生的奥林匹克选手和一个后天需要把自己练运动都做不好,也需要去练习奥林匹克的选手,完全是不一个级别上面的。 他练他的,我练我的

Sunday, June 18, 2023

how to optimize gpt code generation

 i use gpt for optimize my reactjs writing time, i typically write a reactjs objective, let reactjs write code, I wait for it to generate, once generate I diff the generated code with existing code and findout the different and see whether it makes sense, if not, will feedback, if yes, will run it to test it, if ok, then will write commit message and commit the changes. Is there anyway I can optimize this process?


Certainly, your existing process sounds efficient, but here are a few strategies you might consider to further optimize it:

  1. Use Git Hooks: You can use pre-commit and post-commit hooks in Git to automate some of your tasks. For instance, you could set up a pre-commit hook to automatically run your tests and prevent the commit if they fail. A post-commit hook could be used to automatically generate a diff and print it out for you.

  2. Automate Feedback: If you're finding that you're often giving the same feedback or making the same changes, consider writing scripts to automate those changes. You might be able to create a script to analyze the generated code and either automatically make certain changes or give you specific feedback.

  3. Parallelize Your Tasks: Depending on your setup, you might be able to parallelize some of your tasks. For example, while GPT is generating the React code, you might be able to work on other tasks.

  4. Leverage GPT More: If GPT-4 is useful for generating code, it might also be helpful for other tasks. For instance, you might use GPT to help write your commit messages, or even to analyze the diffs.

  5. Use CI/CD Pipeline: You can integrate your process with a CI/CD (Continuous Integration/Continuous Deployment) pipeline. Once you commit your changes, the pipeline can take care of running tests, generating builds, and deploying code to a test environment. This could save you some manual steps.

  6. Use Code Formatter Tools: Use code formatting tools like Prettier for your React code. This will keep your code format consistent and can be set up to run automatically before you make a commit.

  7. Code Review: If you're working with a team, a peer review process can help ensure that changes make sense and that the code is of high quality. Even when working alone, taking a moment to review your code (perhaps after stepping away for a bit) can help you spot issues and improvements.

  8. Optimize GPT Parameters: Make sure you're optimizing the parameters for the GPT model to get the most useful code generation. This could involve fine-tuning the temperature, max tokens, and other parameters.

Remember, the key to optimizing any process is iteration. Try out different methods and see what works best for you, and don't be afraid to change your process if you find a better way of doing things.

诸行无常 一切皆苦 和空有什么关系

 gpt: 诸行无常

一切皆苦 和空有什么关系


"诸行无常"、"一切皆苦"和"空"都是佛教中重要的教义,它们之间存在紧密的关系。

  1. 诸行无常:指的是一切事物都是无常和变化的。万事万物都在不断地生起、存在和灭亡。没有任何事物是永恒不变的,所有的现象都是流转不息的。

  2. 一切皆苦:这是佛教四圣谛(四个基本真理)之一。它指出生命中的存在充满了苦难、痛苦和不满。无论是身体上的疼痛、生老病死的过程,还是心理上的痛苦、迷惑和纠结,一切都被认为是不可避免的苦。

  3. 空:在佛教中,"空"(梵文:śūnyatā)是指一切现象的本性为空性或无我性。它指出一切现象都是没有独立实性、固定实体的。一切事物的存在都是相对和依赖性的,没有永恒、独立、实体的自性。

这三个教义之间的关系可以解释如下:因为一切事物都是无常和变化的(诸行无常),所以它们无法提供持久的满足和安稳。而这种无常性和不确定性带来了苦难和痛苦(一切皆苦)。通过深入领悟事物的无常性和苦难性,我们可以觉悟到事物的真实本质——空性。空性的理解使我们超越对事物的执着和贪爱,最终达到解脱和内在的自由。

总结起来,"诸行无常"提醒我们一切事物都是变化无常的,"一切皆苦"指出存在的痛苦和不满,而"空"则是指出事物的无我性和相对性。这些教义共同帮助人们认识到世界的真实本质,超越痛苦,追求解脱和内在的平静。


诸行无常 一切皆苦 诸法无我 寂灭为乐 诸行无常 是生灭法 生灭灭已 寂灭为乐