Why Software Development is not the same as it was in 2020 (pre-AI era)?

The domain of Software Development / Programming has been in an ever-evolution ever since 1980s when PC came to prominence.

But 2020-2025 might be the biggest leap in Software Development in terms of advancements of all the times.

Of course, due to AI.

Below is a list of some of the tasks where AI is “situationally” saving about 50% – 80% of time:

1. UI Development

User Interface Development is a part of Frontend Development which involves creating user ending interfaces. Basically those parts of website which users directly see.

Before AI

Developers spent days learning CSS, Flexbox, Grid, SCSS, Tailwind, Material UI, etc
Creating a decent-looking UI required strong design intuition
Frontend progress was heavily dependent on experience. less-experiences devs would take very long as compared to senior UI / frontend dev

After AI

AI can generate clean, responsive UI from plain english prompts
Landing pages, dashboards, and basic web apps are created in minutes (optimizing that still takes a few hours tho)
Very less time fighting with styling code and following best best practices, as AI knows syntax and all best practices
UI Devs are essentially on steroids now

Challenges?

Generated UI can look generic or same-ish. (like someone can readily spot most sites being created with AI)
- Methods are being explored to make AI more diverse. Check Claude blog for more
Translating Figma to Code is still something AI struggles with
AI gets sometime stuck at making basic changes like “move this button at top” or “why are these two sections overflowing”

2. Third Party API Integrations

Before AI

Read thousands of words from API docs and carefully select the specific API we need for our use case
Write integration logic and creating clients and handlers in code

After AI

Tell AI about use case, and it can read even hundreds of docs pages, and gets us the APIs we need
Very fast to create integration in existing application code

Challenges?

AI does a terrible job sometimes with hallucinations, and makes up API endpoints which might not exist, potentially misleading
- this issue aggravates for massive API providers like Shopify where AI tends to make some API endpoints out of thin air
Needs to be still properly tested inside the application as AI context is very limited when codebase is large

3. Debugging

Debugging is process of troubleshooting errors arising in code. I personally love this the most.

Before AI

Debugging an issue could easily take hours or days sometimes
Process was Google → StackOverflow → Docs → trial and error
Debugging was most frustrating thing for a developer (still is tho haha)

After AI

Paste errors and relevant code, get explanation instantly
Less time stuck on silly mistakes, and more time spent on productive things
A happier software developer

Challenges?

AI confidently suggests wrong solutions, so one needs to be vigilant
AI knowledge is cut-off so library, SDKs, API docs/links have to be provided manually
Sometimes it takes few mins to manually debug but AI could take hours of iterations. Been there.

4. Documentation or Knowledge Bases

Before AI

Docs were outdated, often missing totally
Very painful and monotonous for software devs to manually write docs (something that is not code)

After AI

AI is remarkable at going through small to mid level codebases, and consolidating code into documentations

Challenges?

AI loves to hallucinate sometimes, and documents things which might not really matter
Since AI bases its judgment based on the knowledge available, a lot of times reasons for writing code in a certain way cannot be inferred directly from code. So human in loop is essential

5. Code Writing As whole

Modern AI models from Claude (clear winner here), OpenAI and Gemini are very good for creating code like junior to mid level software engineer. This has allowed the software developers to augment writing code and shipping features really really fast. A feature which would take 2 months to develop 5 years ago, can now, in most cases, be developed in 2 weeks.

What this means for software industry, specifically for software developers, is that they cannot focus on being good enough only, and should use AI tools as much as possible to speed up their programming process.

We at Barq Dev use AI in a variety of ways, another blog post for that.

This blog mostly focused on bright side of things, I plan to write some other time where AI might not be a good fit for lot of cases. Another blog post for that.