DEV Community: David_Is_Back!

Photoshop in a Figma World: What It's Still the Right Tool For

David_Is_Back! — Thu, 20 Aug 2026 10:22:01 +0000

Photoshop in a Figma World: What It's Still the Right Tool For

Let's get the argument out of the way: you should not be designing app layouts in Photoshop in 2026. Figma won that. Vector-first, real components, collaborative, browser-based, dev mode. If someone hands you a PSD as a spec, that's a smell.

But the pendulum swung too far, and there's now a generation of frontend devs who have never opened Photoshop and hit a wall roughly once a month because of it.

Here's the honest division of labour.

What Figma owns

Layout and spacing
Components and design systems
Prototyping and handoff
Anything a developer needs to inspect
Icons and vector illustration

Don't fight this. If you're doing any of the above in a raster tool, you're making your own life harder.

What Photoshop still owns

1. Anything involving actual photographs

Figma can crop and mask. It cannot retouch.

Product shots for an e-commerce site, hero images, team photos where the lighting is wrong on one person, removing the sign in the background nobody noticed. This is raster work and it needs a raster tool.

The specific skills that matter: masking (especially hair and edges), levels and curves, healing, and non-destructive adjustment layers.

2. The asset export pipeline

This is the one that bites developers most often.

hero.jpg   2.4 MB   ← what the client sent
hero.webp  180 KB   ← what should ship

Photoshop's export handles the things you actually care about:

Format conversion to WebP and AVIF with real quality control
Density variants — @1x, @2x, @3x from one source in one operation
Batch actions — apply the same treatment to 200 product images without opening them individually
Precise quality tuning — where the file size cliff is for this image, not a generic preset

If you've ever shipped a 4 MB PNG because it was "just a background," this is the gap.

3. Texture, compositing, and generated backgrounds

Noise, grain, gradient meshes, duotones, subtle paper textures. Everything that makes an interface look less like a default Tailwind template.

Most of these are raster effects. You can fake some in CSS — and often should, for performance — but designing them means seeing them first.

4. Scripting and automation

Underused by developers, which is odd, because it's the part most aligned with how we think.

Photoshop is scriptable in JavaScript. Resize, watermark, export, rename across a whole directory, driven by code. If you maintain a site with a lot of imagery and you're doing this by hand, you're doing it wrong.

The workflow that actually works

Design in Figma. Layout, components, spacing, the whole interface
Prepare raster assets in Photoshop. Retouch, composite, texture
Place finished assets back into Figma for context
Export from Photoshop, not from Figma, when the output is a photo — you get better format control
Optimize in the build pipeline — Sharp, imagemin, or whatever your framework gives you

Photoshop sits upstream of the design tool, not in competition with it.

What to actually learn

You don't need the whole application. The 20% that covers almost everything a frontend dev needs:

[ ] Layers, groups, and layer masks
[ ] Adjustment layers (non-destructive editing)
[ ] Selection tools — quick select, refine edge
[ ] Smart objects and why they matter for scaling
[ ] Export As / Save for Web, and the format tradeoffs
[ ] Actions and batch processing
[ ] Basic retouching — healing brush, clone stamp

That's maybe fifteen hours of focused learning, and it will pay off for the rest of your career, because photographs are not going anywhere.

Free places to learn it:

Adobe's own tutorials — free, well made, and surprisingly good at the fundamentals
PhotoshopCAFE and Phlearn's free content on YouTube — practical, project-based
Master App & Web Design with Adobe Photoshop A-Z: free and downloadable, aimed at the app and web design use case specifically

!!! Skip anything teaching you to build page layouts in Photoshop. That part of the workflow is genuinely dead — take the raster and asset-pipeline skills and leave the rest.

The takeaway

"Photoshop is obsolete for web design" and "Photoshop is useless" are different claims, and only the first one is true.

Learn Figma for interfaces. Learn enough Photoshop to stop shipping 2 MB heroes.

CompTIA CloudNetX (CNX-001): What It Actually Tests, and Whether You're Ready

David_Is_Back! — Tue, 18 Aug 2026 10:34:16 +0000

There's a certification most people in cloud haven't heard of yet, and the reason is that CompTIA aimed it at a much narrower audience than their usual stuff. Let's talk about what CloudNetX actually is, because the marketing copy around it is uniformly useless.

First: this is not another entry-level cert

CompTIA's reputation is built on Network+, Security+, A+, the on-ramp certifications. CloudNetX is not that.

The recommended background is roughly ten years in IT, five of those as a network architect, plus Network+, Security+ and Cloud+ or equivalent experience.

That's not a gate (nobody checks) but it is a warning. If you take this exam because it looks like a fast credential, you will pay the fee and fail.

What it covers

The premise of the cert is the thing most networking certifications still avoid: almost nobody runs a pure environment anymore. You have on-prem, you have two or three clouds because of an acquisition, you have SD-WAN to branches, and you have remote users who are nowhere. CloudNetX tests the messy middle.

The domains break down roughly as:

Network architecture and design

Hybrid topology planning, IP schema design, subnetting, NAT, and the connectivity fabric between environments.

The practical version of this domain is questions like:

You're connecting an on-prem data centre to two cloud providers. Where does routing live, and what happens when a route is advertised from both sides?
Two business units both use 10.0.0.0/16 internally. You just acquired one of them. Now what?
Transit gateway or peering? At what point does mesh peering become unmanageable?

Overlapping CIDR ranges after a merger is the single most common real-world version of this problem, and it has no clean answer, which is exactly why it makes good exam material.

Security design

Zero Trust architecture, segmentation, firewall policy, secure edge, access control.

The shift being tested here is from perimeter thinking to identity-and-context thinking. If your mental model is still "inside the firewall is trusted," this domain will hurt.

Expect SASE and SD-WAN concepts, east-west traffic inspection, and microsegmentation.

Monitoring and automation

Network visibility, performance tuning, proactive alerting, and infrastructure as code applied to networking.

This is the domain where a lot of traditional network engineers lose points. Networking has been slower to adopt IaC than compute, and "I configure it in the console" is not an answer here. Terraform for network resources, config as code, drift detection.

Troubleshooting

Diagnosing failures that span environments, which is genuinely harder than diagnosing within one, because your tooling changes at every boundary and nobody owns the whole path.

Hybrid DNS resolution is the classic. Forward, conditional forward, private zones, split-horizon, and the failure mode where it works from one subnet and not another.

The exam format


Questions	Up to 90
Format	Multiple choice + performance-based (PBQs)
Duration	165 minutes

The PBQs are the part to prepare for specifically. They're simulated scenarios, not recall. You can know every definition in the objectives and still lose the exam in the PBQs because you've never had to do the thing under time pressure.

Practical advice: do the PBQs last. They eat time unpredictably. Flag them, clear the multiple choice, come back with whatever's left.

Should you take it?

Probably yes if: you're already architecting hybrid networks and you need a vendor-neutral credential that says so. Especially if your experience is spread across AWS and Azure and you're tired of holding two vendor certs that each describe half your job.

Probably no if: you're early career. Network+ then Security+ is a far better use of the same money and time. CloudNetX assumes the fundamentals rather than teaching them.

Maybe if: you're a strong single-cloud engineer moving into a multi-cloud role. The cert's framing, hybrid as the default, not the exception, is genuinely useful even if you never sit the exam.

Free study material

The vendor-neutral angle cuts both ways: there's no single vendor giving away the material, so you assemble it.

CompTIA's official exam objectives PDF. Free, and it's the actual blueprint. Read it before buying anything. Every legitimate course is built from this document
AWS and Azure networking documentation, Transit Gateway, Direct Connect, VPC design, ExpressRoute, Virtual WAN. Free, deep, and it's where the concrete examples come from
The AWS Well-Architected Framework and Azure's equivalent, free PDFs, and the design-tradeoff language maps closely onto how the exam frames questions
CNX-001 Exam Preparation 2026, free structured prep with downloadable material, useful if you want the domains covered in order rather than assembling from docs yourself

The thing that actually gets people through

Build a hybrid lab. Free tier on two clouds, a VPN between them, a routing problem you have to solve. Two weekends of that teaches more than fifty hours of video, because the failure modes are the curriculum.

The exam rewards people who have been broken by hybrid networking before. Go get broken by it cheaply first.

Git Automation from Chat apps

David_Is_Back! — Fri, 07 Aug 2026 09:49:49 +0000

Hey buddies!

It's been a long time since I've wrote anything, but I was creating a free course (the people who know me are aware that I'm always excited to teach) and I'm excited to share it with y'all.

It's not glamorous but it's honest work. Introducing Git Automation from Chat apps.

Who hasn't been outside without a laptop and had some fires to put out? Well, this is for you then. This is the course syllabus:

Managing Git repositories often involves repetitive commands and context switching, especially when teams rely on messaging platforms for communication. This course addresses that by teaching you how to use CrewAI to build intelligent agents that handle Git operations directly from Slack, WhatsApp, or Telegram. You'll see how to reduce manual work and streamline version control tasks.

We'll cover setting up CrewAI, integrating with platform APIs, defining agent tasks for common Git actions, and building practical bots. Includes video lessons, real case studies, and step-by-step guides to apply what you learn.

This is from zero to one course, you'll learn everything you need. Ping me if you have any questions!

Build a Subtitle directory

David_Is_Back! — Thu, 14 Sep 2023 08:43:37 +0000

Why Build a Subtitle Directory Website?

Accessibility: Provide subtitles for videos in various languages to make content accessible to a global audience.

Content Variety: Offer subtitles for movies, TV series, and videos from different sources.

Learning Aid: Subtitles can serve as a valuable tool for language learners.

Now, let's dive into the steps to create your own subtitle directory website.

Step 1: Set Up Your Development Environment

Before you begin, ensure you have Python installed. You'll also need libraries such as Flask for web development and BeautifulSoup for web scraping. You can install these using pip:

pip install Flask beautifulsoup4 requests

Step 2: Design Your Website

Decide on the structure and design of your subtitle directory website. Sketch out the user interface and create HTML templates using a web framework like Flask. You'll need templates for the homepage, search results, and subtitle details.

Step 3: Web Scraping

To scrape subtitles, identify reliable websites or APIs that host subtitles. Some popular sources include Subsfinder, OpenSubtitles, Subscene, or YIFY Subtitles. Using Python's requests library, you can fetch web pages and BeautifulSoup to parse HTML content.

Here's an example of scraping subtitles from a hypothetical website:

`import requests
from bs4 import BeautifulSoup

def scrape_subtitles(movie_name):
url = f'https://example.com/search?q={movie_name}'
response = requests.get(url)

if response.status_code == 200:
    soup = BeautifulSoup(response.content, 'html.parser')
    # Parse the subtitles and return them
    subtitles = soup.find_all('a', class_='subtitle-link')
    return subtitles
else:
    return None

Step 4: Create a Database

You'll need a database to store subtitle metadata, such as movie titles, languages, and download links. SQLite or PostgreSQL are good choices for small to medium-sized projects. Use an Object-Relational Mapping (ORM) library like SQLAlchemy to interact with the database.

Step 5: Build a Search Engine

Implement a search feature on your website that allows users to search for subtitles by movie title or language. Use SQL queries to retrieve relevant subtitle data from your database and display it on your website.

Step 6: Deploy Your Website

Choose a hosting platform, such as Heroku, AWS, or DigitalOcean, to deploy your subtitle directory website. Make sure to configure your environment variables, set up your database, and handle domain and SSL certificates for security.

Step 7: Testing and Maintenance

Thoroughly test your website to ensure it works smoothly. Regularly update your scraping code to adapt to changes on the source websites. Listen to user feedback and make improvements to enhance the user experience.

Congratulations! You've now created a subtitle directory website that can help people access content from around the world. Building such a platform can be a rewarding experience, both technically and socially, as it promotes inclusivity and language learning in the digital age.

Happy coding!