Ege Ersoz

B2B IT consultant by day, Rails enthusiast by night.

[Ege Ersoz @ege02]

@josevalim As the kids say these days: rekt

[Ege Ersoz @ege02]

@PeterDiamandis I actually think it’s closer to $50. All you need is a Claude/GPT subscription to walk you through the business plan and proof of concept product, then file some forms, right?

[Ege Ersoz @ege02]

I've noticed this about myself. Reading AI doomer content makes me pessimistic about our AI future. Then I start building with AI and the possibilities make me excited again.

[Peter H. Diamandis, MD @PeterDiamandis]

Your inputs determine your outputs. If you spend your days consuming fear-based content about AI, you will be afraid. If you spend your days building things with AI, you will be excited. This applies to almost anything.

[Ege Ersoz @ege02]

@steve_hanke Why is Bulgaria in the middle of the Black Sea?

[Ege Ersoz @ege02]

I'm more cynical, Peter. Both Sam and Dario walked back their apocalyptic predictions because they were pulled aside and told that it makes them and their products extremely unpopular and just fans the flames of anti-AI sentiment. Considering both companies want to IPO soon, they obviously don't want that.

[Ege Ersoz @ege02]

@bryan_johnson What about women?

[Ege Ersoz @ege02]

@rickasaurus Not if their salaries were $200/mo. Or even $1,000/mo.

[Ege Ersoz @ege02]

@cgarciae88 "Online commentary" = this website? Because that's not what I'm seeing on other places at all.

[Interesting things @awkwardgoogle]

The perfect wallet

[Ege Ersoz @ege02]

People love to point out that AI is good at coding, and the skeptics always fire back with “oh yeah? then where are all the profitable startups made by AI???”

Two reasons you don’t see them:

1.Most of those founders aren’t engineers. They’re not hanging around Hacker News doing Show HN posts to flex.

2.The technical ones have no reason to talk. When AI can clone any app in a day, announcing your profitable idea to the world is the last thing you’d do.

[Ege Ersoz @ege02]

@PeterDiamandis Agreed. Most people are bad at math and don’t have intuitive grasp of exponential growth. Even the ones who think things are starting to happen too fast are in for a rude awakening!

[Ege Ersoz @ege02]

@PeterDiamandis Yes. AI feels like a superpower, especially when you configure it properly. "Tireless cofounder" is the perfect way to describe it.

[Ege Ersoz @ege02]

Great tip! I also use /grill-me skill from @mattpocockuk

[Thariq @trq212]

been asking others at Anthropic how they stay in the loop with Claude and fully understand the work being done

this is one of my favorites from Suzanne:

[DHH @dhh]

"This is a protectionist tale as old as time. And the justifications are just as tired: It's about quality! It's about attribution! It's about workers! Spare me. It's about you, your insecurities, and your privileges." https://t.co/SP6DubrXXh

[Ege Ersoz @ege02]

Counterpoint: the 1st gen I used for years back in the day before stashing into a cabinet still works today. So does my 2010 MacBook Air. Cannot say the same about any other non-Apple device I’ve ever owned. High-end Android devices various members of my family use tend to last less than five years.

[Ege Ersoz @ege02]

Sorry Peter, I'm an optimist but I don't fully buy this. A hundred years ago people envisioned ubiquitous flying cars and jetpacks. Today, we have helicopters (rare) and no jetpacks. Just because we're imagining something about the future doesn't necessarily mean it will happen. In fact go look at old science/industry magazine covers and you'll see things that were pure fantasy back then and still is pure fantasy today.

[Ege Ersoz @ege02]

@dhh We're starting to see a grand bifurcation in software development. But ultimately it will be no more meaningful than organic vs. regular produce.

[Ege Ersoz @ege02]

@PeterDiamandis I’m a software engineer. I have friends who have never used AI and refuse to. They think it rots their brain and causes skill atrophy. Makes no sense to me. It’s like insisting on riding on horseback when cars have been invented.

[Ege Ersoz @ege02]

Like many, I use both Claude and Codex. But Codex has been annoying me a lot lately.

Opus 4.8 was working on something when it found a somewhat obscure issue in the API for Feature A and suggested hardening it. Out of curiosity, I asked Codex 5.5 how it would fix it. Codex investigated, verified the finding and wrote a plan.

I gave the plan to Opus. It looked at the plan, did a bunch of digging, then said “the proposed fix makes sense in a vacuum, but I see that features B and C are tied to Feature A in such and such way, which makes me think the fix would break the intended design and result in regression in X, Y and Z use cases”.

I looked at the code and realized Claude was 100% correct. I then had it write a new plan and implement it.

Had I trusted Codex, the issue would have been "fixed", but users relying on those features for their use cases would have experienced regressions.

Don't get me wrong: Codex is pretty good at surgical fixes or following a plan, which is good for when you know exactly what needs to be implemented and how. But it only does what it's told and no more. Opus is a better programming companion who doubles as a product manager and domain expert. It has strong product sense and good design taste. People complain about how it burns a lot of tokens, but in my experience it's well worth it.