No AI used in this post. Other posts may contain up to 10% AI assistance.

It’s been nearly three years since SWEs started actively using coding agents.

The claims around Coding Agents in SaaS development:

  • Products will be developed faster
  • Single engineers will cover more verticals
  • Engineers will become 10x within their verticals
  • Companies will profit faster
  • Startups will fail and adapt faster
  • Teams will become smaller with less friction

are kinda correct, but not as exaggerated. Why?

I’ll actually try to answer two questions:

  1. Why don’t we see more SaaS Products since Coding Agents?
  2. Engineers code 10x faster with agents, completing MVP in 2 months. If no agents, would’ve taken 20 months?

First of all, why my ideas might matter

For the last 6 years, I’ve been lucky enough to work at startups and corporates developing AI SaaS products, both B2B and B2C, mostly from day one. I’ve coded hundreds of thousands of lines before coding agents, which taught me how to become a better programmer. Coding Agents later taught me how to become faster. I’ve been lucky enough to heavily contribute to all aspects of product development, engineering, design, demos, fights. Evidence, one of them, career.io/interview-prep, AI product helping users prepare for any interview.

Service for stalkers, my Linkedin Account.

What has changed in the product development since Coding Agents

  • Engineers and Product Managers deliver more and more nice-to-have (I don’t want to say sh itty) features compared to Cursor era. The path to profit still depends on value proposition, PMF, and marketing/sales/SEO teams.
  • Must-have features, and core (of the) product, that expects to drive revenue, has always been developed first. It doesn’t really matter if you release in 1 month or 4, not every competitor is chasing you.
  • Even though there are competitors chasing you, you are all moving at roughly the same development speed and you’ll share the pie rather than one of you dominating the market.

Non-SWE reasons

  • Profitable ideas are harder to find, not engineers, not AI-Agents. I’ve worked for an employer earning lots from its first product because it was a huge PMF, all other products developed afterwards were just, meh. Companies mostly have one-good-sold-idea, or a single PMF, and ideas are not found by Coding Agents.
  • More competitors in domain, more and more confusion for potential customers (both for B2B and B2C). Any service you’re actually willing to pay for, or any business idea, just search it on Google and you’ll see first 2 pages is nothing but different companies offering similar products, harder to choose.
  • Evidence? Search ‘interview prep’ on Google and see how many AI Interview Preparation apps, including the one I developed (its backend and ai).
  • Startups are more flexible to change product directions, that doesn’t mean CEOs/PMs are flexible to redirect towards more profitable paths. Claude Cowork didn’t make them more intelligent, or if it has done, everyone has equally upskilled.
  • It’s always about the sales, not the engineering team quality. Even though engineering teams are skillfully 2-3x minimum, it leads nowhere when sales, PMF.
  • Startups mostly have one year of run-rate to release the core product. If a startup releases the MVP in 1 month or 4 months(prior to AI Agents), there’s no significant change in their life. That’s the main idea on seed-round right? You are buying time and money to test how profitable the core product is. The rest is, “can you achieve to PMF?”. Engineers have gifted 3 additional months to leadership team, but 3 extra months alone doesn’t guarantee profitability.

Engineers code 10x faster with agents, completing MVP in 2 months. If no agents, MVP would’ve taken 20 months?

  • Engineers Cursor, PMs using Claude Cowork, finally a startup having an MVP in 2 months.
  • It’s obvious that you’ve written 10x code, PM used Claude Cowork, LP used ChatGPT, everyone is at 10x. It’s not a lie.
  • Does that mean, it would have taken nearly 2 years to MVP in 2020? No, a startup wouldn’t have sufficient funding of two years, or a corporate company wouldn’t have a patience.
  • Probably it would take around 6 months. So products are being developed 3x faster, not 10x.

What steals the remaining 7x?

  • MVP’s were ok to buggy. Nowadays startups/companies are perfecting their MVPs.
  • MVP’s were ‘minimum’, with less nice-to-haves, now the barrier is higher, companies/startups developing more features, and mostly nice-to-haves.
  • Leadership Teams execute the all ideas due to FOMO, not validating the market first.

Why Coding Agents Haven’t Revolutionised SaaS as its been told, yet

  • Prior to Coding Agents, a SWE manager/lead/CTO was kinda experiencing Cursor-like engineering, companies with sufficient budget have always attacked for the potential market. So nothing new in here.
  • As an engineer/team, your skills 3x’ed or 10x’ed. That doesn’t mean every infrastructure or marketing decision suddenly makes the business value proposition 3x–10x stronger, may be weaker.
  • Upskilling means you have now more skills. It means more brainrot.
  • You are not the bottleneck between AI (or coding) Agents and correct/faster decisions. You are the orchestrator that have to be there.

Who has benefitted from Coding Agents then?

  • Corporates. It’s for sure that AI upskilled engineers, unfortunately less fellow engineers are required to maintain existing profitable solutions. Layoffs are nothing but expected.