I compared myself with current gen AI and reviewed them

Jason Voorhees

Jason Voorhees

𝕯𝖝𝕯 𝖈𝖗𝖊𝖜 𝕵𝖊𝖘𝖙𝖊𝖗
Joined
May 15, 2020
Posts
61,340
Reputation
172,412
I’ve been writing code for 5 years now, and over the past few weeks, I’ve been extensively using a bunch of large language models Claude, GPT-4.5, DeepSeek, LLaMA, Gemini, etc across both work and personal projects.

And I’ll be real for many tasks, especially small-to-medium ones, they’re faster, more consistent, and make fewer mistakes than I do. That’s just the truth. But in real world projects, human developers still have a place and I still think devs like me bring a lot to the table.


Claude Opus (Anthropic – Paid)

A calm, thoughtful generalist. Structured, logical, and great for thinking through architecture and code flow.

Beats me at:

Writing clean code on the first try

Refactoring and reviewing code

Thinking through edge cases clearly

Generating hours of work in seconds



GPT-4.5 (OpenAI via ChatGPT Plus)

The all-rounder. Whether it’s frontend (React), backend (Node), database (Mongo), or testing — it handles it all.

Beats me at:

Generating boilerplate and repetitive code

Writing test cases and documentation

Prototyping full-stack UIs quickly



DeepSeek R1

Basically an engineer's assistant. Especially good with ML workflows, math-heavy tasks, and algorithmic logic.

Beats me at:

Writing efficient ML code with minimal prompt

Solving stats-heavy and algorithm-based problems

Competitive programming-level logic puzzles


Claude Sonnet (Free)

The free-tier surprise. Handles day-to-day scripting and simple development work really well.

Beats me at:

Bash, SQL, scripting small utilities

Explaining code and suggesting fixes

Handling small tasks with minimal input



Code LLaMA 70B / LLaMA 3 (Meta – Open Source)

Self-hostable and very capable. Shines with low-level languages and obscure syntax.

Beats me at:

C++, Rust, Go, and niche language support

Understanding obscure syntax better than I do sometimes

Writing detailed, well-commented code


Gemini 1.5 Pro (Google)

Context window king. It can take in entire codebases, UI specs, docs, and spreadsheets all at once.

Beats me at:

Understanding and summarizing huge codebases

Spotting inconsistencies across large projects

Integrating across docs, code, APIs, and design specs



So is it over?

If you’re building a one-off CRUD app? A Bash script? A niche ML tool?
Then yes. It is over, you probably don’t need me.

These models can:

Build and deploy full-stack apps

Automate workflows

Solve complex math/programming problems

But in long-term projects, with shifting goals, unclear requirements, messy integrations, human clients, and real-world edge cases? You still need a developer.Because human devs can:

Understand nuance and client intent

Make trade-offs and prioritization decisions

Adapt on the fly when things go off-script

Handle team dynamics and real-time feedback

Think creatively in weird, domain-specific edge cases

TLDR-

Yes, they do mog me hard in many things. They faster, more accurate and smarter than me in many places But not everywhere atleast not yet. I also used AI to write this thread if you couldn't tell btw. Jfl
 
Last edited:
  • +1
  • So Sad
Reactions: Depresso, MiserableMan, loyolaxavvierretard and 9 others
@loyolaxavvierretard @deadstock @Swarthy Knight @Debetro @optimisticzoomer @gooner23
 
  • +1
Reactions: MiserableMan, loyolaxavvierretard, aladdinmaxxer and 2 others
I’ve been writing code for 5 years now, and over the past few weeks, I’ve been extensively using a bunch of large language models Claude, GPT-4.5, DeepSeek, LLaMA, Gemini, etc across both work and personal projects.

And I’ll be real for many tasks, especially small-to-medium ones, they’re faster, more consistent, and make fewer mistakes than I do. That’s just the truth. But in real world projects, human developers still have a place and I still think devs like me bring a lot to the table.


Claude Opus (Anthropic – Paid)

A calm, thoughtful generalist. Structured, logical, and great for thinking through architecture and code flow.

Beats me at:

Writing clean code on the first try

Refactoring and reviewing code

Thinking through edge cases clearly

Generating hours of work in seconds



GPT-4.5 (OpenAI via ChatGPT Plus)

The all-rounder. Whether it’s frontend (React), backend (Node), database (Mongo), or testing — it handles it all.

Beats me at:

Generating boilerplate and repetitive code

Writing test cases and documentation

Prototyping full-stack UIs quickly



DeepSeek R1

Basically an engineer's assistant. Especially good with ML workflows, math-heavy tasks, and algorithmic logic.

Beats me at:

Writing efficient ML code with minimal prompt

Solving stats-heavy and algorithm-based problems

Competitive programming-level logic puzzles


Claude Sonnet (Free)

The free-tier surprise. Handles day-to-day scripting and simple development work really well.

Beats me at:

Bash, SQL, scripting small utilities

Explaining code and suggesting fixes

Handling small tasks with minimal input



Code LLaMA 70B / LLaMA 3 (Meta – Open Source)

Self-hostable and very capable. Shines with low-level languages and obscure syntax.

Beats me at:

C++, Rust, Go, and niche language support

Understanding obscure syntax better than I do sometimes

Writing detailed, well-commented code


Gemini 1.5 Pro (Google)

Context window king. It can take in entire codebases, UI specs, docs, and spreadsheets all at once.

Beats me at:

Understanding and summarizing huge codebases

Spotting inconsistencies across large projects

Integrating across docs, code, APIs, and design specs



So is it over?

If you’re building a one-off CRUD app? A Bash script? A niche ML tool?
Then yes. It is over, you probably don’t need me.

These models can:

Build and deploy full-stack apps

Automate workflows

Solve complex math/programming problems

But in long-term projects, with shifting goals, unclear requirements, messy integrations, human clients, and real-world edge cases? You still need a developer.Because human devs can:

Understand nuance and client intent

Make trade-offs and prioritization decisions

Adapt on the fly when things go off-script

Handle team dynamics and real-time feedback

Think creatively in weird, domain-specific edge cases


So yeah, they do mog me hard sometimes. They faster, more accurate and smarter than me in many places But not everywhere atleast not yet.
damn
 
  • +1
Reactions: MiserableMan, aladdinmaxxer, truthhurts and 1 other person
@imontheloose
 
  • +1
Reactions: MiserableMan, aladdinmaxxer, truthhurts and 1 other person
brutaly over for CS majors
 
  • +1
Reactions: MiserableMan, aladdinmaxxer, truthhurts and 1 other person
Dnr
 
  • +1
Reactions: aladdinmaxxer, Jason Voorhees and truthhurts
Post made by AI btw.
 
  • +1
  • JFL
Reactions: MiserableMan, aladdinmaxxer, browncurrycel and 2 others
Dnr
 
  • +1
Reactions: Jason Voorhees
I’ve been writing code for 5 years now, and over the past few weeks, I’ve been extensively using a bunch of large language models Claude, GPT-4.5, DeepSeek, LLaMA, Gemini, etc across both work and personal projects.

And I’ll be real for many tasks, especially small-to-medium ones, they’re faster, more consistent, and make fewer mistakes than I do. That’s just the truth. But in real world projects, human developers still have a place and I still think devs like me bring a lot to the table.


Claude Opus (Anthropic – Paid)

A calm, thoughtful generalist. Structured, logical, and great for thinking through architecture and code flow.

Beats me at:

Writing clean code on the first try

Refactoring and reviewing code

Thinking through edge cases clearly

Generating hours of work in seconds



GPT-4.5 (OpenAI via ChatGPT Plus)

The all-rounder. Whether it’s frontend (React), backend (Node), database (Mongo), or testing — it handles it all.

Beats me at:

Generating boilerplate and repetitive code

Writing test cases and documentation

Prototyping full-stack UIs quickly



DeepSeek R1

Basically an engineer's assistant. Especially good with ML workflows, math-heavy tasks, and algorithmic logic.

Beats me at:

Writing efficient ML code with minimal prompt

Solving stats-heavy and algorithm-based problems

Competitive programming-level logic puzzles


Claude Sonnet (Free)

The free-tier surprise. Handles day-to-day scripting and simple development work really well.

Beats me at:

Bash, SQL, scripting small utilities

Explaining code and suggesting fixes

Handling small tasks with minimal input



Code LLaMA 70B / LLaMA 3 (Meta – Open Source)

Self-hostable and very capable. Shines with low-level languages and obscure syntax.

Beats me at:

C++, Rust, Go, and niche language support

Understanding obscure syntax better than I do sometimes

Writing detailed, well-commented code


Gemini 1.5 Pro (Google)

Context window king. It can take in entire codebases, UI specs, docs, and spreadsheets all at once.

Beats me at:

Understanding and summarizing huge codebases

Spotting inconsistencies across large projects

Integrating across docs, code, APIs, and design specs



So is it over?

If you’re building a one-off CRUD app? A Bash script? A niche ML tool?
Then yes. It is over, you probably don’t need me.

These models can:

Build and deploy full-stack apps

Automate workflows

Solve complex math/programming problems

But in long-term projects, with shifting goals, unclear requirements, messy integrations, human clients, and real-world edge cases? You still need a developer.Because human devs can:

Understand nuance and client intent

Make trade-offs and prioritization decisions

Adapt on the fly when things go off-script

Handle team dynamics and real-time feedback

Think creatively in weird, domain-specific edge cases

TLDR-

Yes, they do mog me hard in many things. They faster, more accurate and smarter than me in many places But not everywhere atleast not yet.
Just a matter of time till devs become close to opsolete probably.
 
  • +1
Reactions: MiserableMan, aladdinmaxxer, Jason Voorhees and 1 other person
Ironic that bro also used AI to make this post
 
  • +1
  • JFL
Reactions: Jason Voorhees and truthhurts
Brutal read:feelsbadman:
 
  • +1
Reactions: Jason Voorhees
it was over for software devs from 2023
 
  • +1
Reactions: MiserableMan, truthhurts and Jason Voorhees
Just a matter of time till devs become close to opsolete probably.
I actually think devs are the last people to lose their jobs. The ones that will lose their jobs first are people doing repetitive Redundant work like accounts, finance etc. AI is already better than all of them @imontheloose
 
Last edited:
  • +1
  • Woah
Reactions: MiserableMan, widdi, pajjeetslayer and 2 others
@sigma boii @SecularIslamist @aladdinmaxxer
 
  • +1
Reactions: truthhurts
Come back to me when someone doesn't know how to code can create apps, fully functional complex scripts etc. I asked deepseek to create a python script for me and it doesn't even work. Also since you're a dev, which area would you recommend people go into to not get AI'D. I'm thinking stuff like dev ops and cloud practitioners are safe, what else though?
 
Last edited:
  • +1
Reactions: Jason Voorhees
Come back to me when someone doesn't know how to code can create apps, fully functional complex scripts etc. I asked deepseek to create a python script for me and it doesn't even work. Also since you're a dev, which area would you recommend people go into to not get AI'D. I'm thinking stuff like dev ops and cloud practitioners are safe, what else though?
Like you said devops, cloud architects, cybersec, embedded systems etc. Will take a while for AI to catch up
 
Come back to me when someone doesn't know how to code can create apps, fully functional complex scripts etc. I asked deepseek to create a python script for me and it doesn't even work. Also since you're a dev, which area would you recommend people go into to not get AI'D. I'm thinking stuff like dev ops and cloud practitioners are safe, what else though?
Ai is still great at code generation tho. Like I made my first app 3 years ago and it took more me almost 3 weeks.


With AI you can recreate this in within 4-5 days easily. A complete noob maybe not but anyone who understands a little about web dev basics and frameworks can totally do this with AI no problem
 
  • +1
  • Hmm...
Reactions: MiserableMan, aladdinmaxxer and JohnDoe
@Pakicel
 
  • +1
Reactions: Pakicel
The repeated use of questions clued me in to the use of AI in the thread
 
  • +1
Reactions: Fusionxz
I’ve been writing code for 5 years now, and over the past few weeks, I’ve been extensively using a bunch of large language models Claude, GPT-4.5, DeepSeek, LLaMA, Gemini, etc across both work and personal projects.

And I’ll be real for many tasks, especially small-to-medium ones, they’re faster, more consistent, and make fewer mistakes than I do. That’s just the truth. But in real world projects, human developers still have a place and I still think devs like me bring a lot to the table.


Claude Opus (Anthropic – Paid)

A calm, thoughtful generalist. Structured, logical, and great for thinking through architecture and code flow.

Beats me at:

Writing clean code on the first try

Refactoring and reviewing code

Thinking through edge cases clearly

Generating hours of work in seconds



GPT-4.5 (OpenAI via ChatGPT Plus)

The all-rounder. Whether it’s frontend (React), backend (Node), database (Mongo), or testing — it handles it all.

Beats me at:

Generating boilerplate and repetitive code

Writing test cases and documentation

Prototyping full-stack UIs quickly



DeepSeek R1

Basically an engineer's assistant. Especially good with ML workflows, math-heavy tasks, and algorithmic logic.

Beats me at:

Writing efficient ML code with minimal prompt

Solving stats-heavy and algorithm-based problems

Competitive programming-level logic puzzles


Claude Sonnet (Free)

The free-tier surprise. Handles day-to-day scripting and simple development work really well.

Beats me at:

Bash, SQL, scripting small utilities

Explaining code and suggesting fixes

Handling small tasks with minimal input



Code LLaMA 70B / LLaMA 3 (Meta – Open Source)

Self-hostable and very capable. Shines with low-level languages and obscure syntax.

Beats me at:

C++, Rust, Go, and niche language support

Understanding obscure syntax better than I do sometimes

Writing detailed, well-commented code


Gemini 1.5 Pro (Google)

Context window king. It can take in entire codebases, UI specs, docs, and spreadsheets all at once.

Beats me at:

Understanding and summarizing huge codebases

Spotting inconsistencies across large projects

Integrating across docs, code, APIs, and design specs



So is it over?

If you’re building a one-off CRUD app? A Bash script? A niche ML tool?
Then yes. It is over, you probably don’t need me.

These models can:

Build and deploy full-stack apps

Automate workflows

Solve complex math/programming problems

But in long-term projects, with shifting goals, unclear requirements, messy integrations, human clients, and real-world edge cases? You still need a developer.Because human devs can:

Understand nuance and client intent

Make trade-offs and prioritization decisions

Adapt on the fly when things go off-script

Handle team dynamics and real-time feedback

Think creatively in weird, domain-specific edge cases

TLDR-

Yes, they do mog me hard in many things. They faster, more accurate and smarter than me in many places But not everywhere atleast not yet. I also used AI to write this thread if you couldn't tell btw. Jfl
What's AI for writing code like? I've always wanted to get into game design and I'm a good texture maker and 3D modeler but I can't code other than some basic script kiddying
 
The future is looking certainly bleak for a lot professions. Sooner or later Ai will take over most jobs, well i hope we can get more free benefits in the future thanks to this, and spend time doing other activities. What do you think?
 
  • +1
Reactions: Jason Voorhees
Yeah. These AI tools are suprisingly good for 95% of use cases. But it's the 5% for which you need a real human in the loop.

Also, has very minimal knowledge of the domain you are working in (including tonds of tribal knowledge that might be in a few people's heads), which is critical to the vast majority of codemonkey jobs tbh.

I still think the days of most SWEs are numbered, and most who aren't white chadlites with trust funds should keep the rope ready imo. But I am confident you will ascend hard in your career. Without some serious blackpilling, WITCH currycels are doomed unfortunately.
 
  • +1
Reactions: Jason Voorhees

Similar threads

D
Replies
20
Views
12K
carlos72
carlos72
D
Replies
29
Views
14K
clover
clover

Users who are viewing this thread

Back
Top