I just watched an AI agent take a Jira ticket, understand our codebase, and push a PR in minutes and

[human1234]

I just watched an AI agent take a Jira ticket, understand our codebase, and push a PR in minutes and I’m genuinely scared

I’m a professional software engineer, and today something happened that honestly shook me. I watched an AI agent, part of an internally built tool our company is piloting, take in a small Jira ticket. It was the kind of task that would usually take me or a teammate about an hour. Mostly writing a SQL query and making a small change to some backend code.

The AI read through our codebase, figured out the context, wrote the query, updated the code, created a PR with a clear diff and a well-written description, and pushed it for review. All in just a few minutes.

This wasn’t boilerplate. It followed our naming conventions, made logical decisions, and even updated a test. One of our senior engineers reviewed the PR and said it looked solid and accurate. They would have done it the same way.

What really hit me is that this isn’t some future concept. This AI tool is being gradually rolled out across teams in our org as part of a pilot program. And it’s already producing results like this.

I’ve been following AI developments, but watching it do my job in my codebase made everything feel real in a way headlines never could. It was a ticket I would have knocked out before lunch, and now it’s being done faster and with less effort by a machine.

I’m not saying engineers will be out of jobs tomorrow. But if an AI can already handle these kinds of everyday tickets, we’re looking at serious changes in the near future. Maybe not in years, but in months.

Has anyone else experienced something similar? What are you doing to adapt? How are you thinking about the future of our field?

 

https://www.reddit.com/r/cscareerquestions/s/TX7b6Ai6D2

[human1234]

 

[Mr Mirchi]

Bongule

antha hadavidi

[futureofandhra]

2 hours ago, human1234 said:

I just watched an AI agent take a Jira ticket, understand our codebase, and push a PR in minutes and I’m genuinely scared

I’m a professional software engineer, and today something happened that honestly shook me. I watched an AI agent, part of an internally built tool our company is piloting, take in a small Jira ticket. It was the kind of task that would usually take me or a teammate about an hour. Mostly writing a SQL query and making a small change to some backend code.

The AI read through our codebase, figured out the context, wrote the query, updated the code, created a PR with a clear diff and a well-written description, and pushed it for review. All in just a few minutes.

This wasn’t boilerplate. It followed our naming conventions, made logical decisions, and even updated a test. One of our senior engineers reviewed the PR and said it looked solid and accurate. They would have done it the same way.

What really hit me is that this isn’t some future concept. This AI tool is being gradually rolled out across teams in our org as part of a pilot program. And it’s already producing results like this.

I’ve been following AI developments, but watching it do my job in my codebase made everything feel real in a way headlines never could. It was a ticket I would have knocked out before lunch, and now it’s being done faster and with less effort by a machine.

I’m not saying engineers will be out of jobs tomorrow. But if an AI can already handle these kinds of everyday tickets, we’re looking at serious changes in the near future. Maybe not in years, but in months.

Has anyone else experienced something similar? What are you doing to adapt? How are you thinking about the future of our field?

 

https://www.reddit.com/r/cscareerquestions/s/TX7b6Ai6D2

Legacy app cheyamanu chudham

[Spartan]

ninnane ma code base ki integration tests rasivva mante...mottam all endpoint with multiple scenarios rasi ichindi...

ma qa gallaki rayandi ra ante...manual meeda focus chesta..a tarvata update chesta ani kathal mingaru 2 sprints... 

monday pipeline lo integrate chesi dobbali.....all i had to do was run `npx run tests --env=dev` and pass env vars.

bleow Prompt:

Generate Integration tests for this repo pointing to dev server. Setup environment variables as suited.

 

 

above case in the article is pretty much like what I did, additional is JIRA integration with pilot.

Prompt would be...pickup this JIRA ticket and update my code for acceptance criteria....

[enigmatic]

5 minutes ago, Spartan said:

bleow Prompt:

Generate Integration tests for this repo pointing to dev server. Setup environment variables as suited.

emi llm vaadaru - copilot or cursor ai or claude ?

 

[Spartan]

1 minute ago, enigmatic said:

emi llm vaadaru - copilot or cursor ai or claude ?

 

copilot ..

cursor enable chesaru..should start using that. 

[Teluguredu]

I already told you ,in future they will integrate LLM's directly into project build.

[A1startarak]

29 minutes ago, Spartan said:

copilot ..

cursor enable chesaru..should start using that. 

Sometimes copilot doesn’t understand modified wrapper functions within. initially it was not accurate for me ,agree that it got much better after 1year of feeding it understands most of repo. Chat kante Edit feature masth undi. Not only analyzing but giving accurate fixes as time goes by. Most useful feature is codereview option👍. Can you tell where is this cursor option ? Is this new plugin?

[A1startarak]

2 hours ago, Mr Mirchi said:

Bongule

antha hadavidi

Layoff ayyaka vachi kanapadu 

[desiboys]

AI is overhyped..don't fall into that trap ani openai and meta have laidoff many AI  employees.

[Assam_Bhayya]

4 hours ago, human1234 said:

 

. . . . . . .created a PR with a clear diff and a well-written description, and pushed it for review. All in just a few minutes.

. . . . .

 

https://www.reddit.com/r/cscareerquestions/s/TX7b6Ai6D2

[Konebhar6]

I have been working on AI and used different LLMs. It’s amazing how far it has come. You gotta learn how to instruct AI via good Prompt engineering. You can easily build AI agents via semantic layer and let them do their magic. 
Also learnt AI can be lazy (wants to get to decision tree quickly. Your questions should be straight forward), acts like a child (you have to instruct clearly). 
But it’s amazing how far it has come. We did customer segmentation and hot 98% accuracy within 2 months. 

 

 

[Konebhar6]

I can write basic python code. I can understand it. I have used chatGPT to write all code I want with all wrinkles and complex logic and it writes excellent. Not even a single mistake. 

[anna_gari_maata]

The dangerous thing in this AI is if we give code base and if AI can read it in a jiffy ... that's a cause for concern 

If AI can read any code base in a jiffy, SDE marchipovacchu