David Vartanian

Founder @ Beamer. I write about engineering cost after PMF and how software companies scale profitably.

[David Vartanian @BeamerSoftware]

https://t.co/WzCNmUrRCT

[David Vartanian @BeamerSoftware]

Yes, but I'm not sure that's accurate.
It's what I experienced: I was using grok bot and was able to make it to day 5 or 6 before the weekly reset.
But now the same level of work uses around 80% of "usage" in one single day, while I was promised "generous "usage" and 2x for Cursor", but it was all bullshit.
So I created my own bot roster based on Hermes bot mode.

[David Vartanian @BeamerSoftware]

@tonysimons Paying the $100 subscription doesn't make any difference.

[David Vartanian @BeamerSoftware]

@aipulseda1ly 2 days left? what a lazy ass. It should've been 5 days left if you really wanted to get shit done.

[David Vartanian @BeamerSoftware]

@elonmusk @bot I tried. I ran out of "usage".
Bye.

[David Vartanian @BeamerSoftware]

@elonmusk @bot It was dialing but it got out of "usage" after the 4th digit.

[David Vartanian @BeamerSoftware]

@nvandewetering I'm replicating all grok bot features using hermes in bot mode.
Besides the nice UI and that it's faster replying messages, there's nothing grok bot has that can't be achieved without it.

[David Vartanian @BeamerSoftware]

Companies will need to invest in understanding the solutions built with AI to take real advantage.
That makes sense to me.

[Guillermo Flor @guilleflorvs]

Mark Cuban's take on the next job wave deserves more attention than it's getting.

His argument: software as we know it is dying. Generic SaaS gets replaced by AI customized to how each business actually operates.

The question then becomes, who builds and deploys that for the 33 million companies in the US that can't do it themselves?

That's the market. Not AI for enterprises with 500-person eng teams. The other 99%.

SMBs don't have CTOs. They don't have prompt engineers. They need someone to sit across the table, understand their workflow, and wire up an AI system that fits it.

That's a services business. Probably the largest one this decade.

We broke this down step by step in Product Market Fit: https://t.co/a6d1HT1c7r

[David Vartanian @BeamerSoftware]

@sabben Decision latency is the one I'd track first. Building got cheap, but work still sits waiting for reviews and sign-offs, and AI doesn't shrink that wait.

[David Vartanian @BeamerSoftware]

@lewang_teraliq In software teams the gain often doesn't reach anyone. Time saved writing code goes into reviewing more of it and fixing what it broke, so output looks bigger and delivery stays the same.

[David Vartanian @BeamerSoftware]

@jamesjansson @ProtonMail Writing the code was never the expensive part. Reviewing it, fixing what breaks and changing it a year later is, and none of that gets cheaper when the devs stop understanding the code.

[David Vartanian @BeamerSoftware]

A pricing change that should take one line ends up touching eleven files in three services. Two reviewers spend the afternoon working out why each file is involved. It ships, and the next day a report nobody connected to pricing stops adding up. Most of that time was paid for by how the code is split, and the change itself was the small part.

https://t.co/QKgCQYCqDl

[David Vartanian @BeamerSoftware]

Here's the 18th article around The Engineering Tax.
This time it goes quite nerdy with all those formulas but how else will you calculate cost other than doing math?

One number that turns a guess into an economic decision. https://t.co/08wN16Sb6T

[David Vartanian @BeamerSoftware]

@brita_lisa Until the old versions age out, the team keeps both paths alive and has to test and support each one on every release. That's engineering time spent on users who haven't updated, and it rarely shows up in the plan.

[David Vartanian @BeamerSoftware]

@JDSalbego @github Each of those agent PRs still lands on a reviewer who has to rebuild how that part of the system fits together before approving it. When one change spreads across many files, review time grows with it.

[David Vartanian @BeamerSoftware]

@spcxGOD @elonmusk @DogecoinNorway You're right but that's unrealistic.
If there's another option for a much lower price people might just cancel and buy something else.

[David Vartanian @BeamerSoftware]

@elonmusk @DogecoinNorway Finally you admit it.

It would be great to make token availability transparent instead of advertising "generous usage" and then it turns out it's worse.

[David Vartanian @BeamerSoftware]

@coreyepstein @HQForWork I disagree, but who cares? That wasn't even the topic.

[David Vartanian @BeamerSoftware]

@coreyepstein @HQForWork Are you saying your product doesn't need to evolve over years? Can you explain what kind of product you maintain?

[David Vartanian @BeamerSoftware]

Anymore never means forever.
And I doubt an agent can create stable durable code that doesn't need any improvement.
But let's say agents do that because that's expected as they keep improving their quality.
I wouldn't worry if engineers don't know the code anymore, but I would worry if no one knows what their product does because market and customers do evolve and eventually that stable software that has no technical needs might still need a change because the demand requires the product to evolve.

[David Vartanian @BeamerSoftware]

@svpino Bug fixes get slow when you have to trace why a billing change broke the export, because both read the same table in different ways. Faster code review doesn't shorten that time. Does anyone on your team measure it?