i’m back

i hope you enjoyed the trailer. i’m currently determining what the real launch video should be. i’m thinking the god’s plan music video or a kanye super bowl ad (or this one). lmk!

hey there! to be honest, i’ve been trying to write this memo for the last 2 days.

i’m buried in a bunch of concepts/ideas, but it’s proved harder than expected to go from a schizophrenic apple note → a nice memo.

however, it’s 1am and i really want to go to bed, so i’m just going to freestyle (haha) it.

the best things in life are free

i’ve been building consumer ai since the davinci-002 days. i can’t say it’s ever really worked, but i’ve had a blast doing it. bringing the most powerful technology of my lifetime to ordinary people is my mission.

my favorite part of the process has always been the 0→1. however, every time i started getting people to pay for it, all the fun got sucked out. freecode is the perfect product for me:

• there is no pricing page or paywall to stress over.
• i don’t feel guilt for not delivering the most capable product out there. i have no problem with users “graduating” to $200 plans.
• a world where access to intelligence is expanded is a better place.
• growth is less of a problem than making sure we have sustainable economics.
• i am super excited to solve this problem. after all, my favorite book is freakonomics.

one fish, three meals

there’s plenty of reasons why my last company fell apart. i think the core issue was the awkward middle ground we were trying to fill.

on one side, they have to compete with the built-in, default, free version inside their browser/phone OS/chat app. on the other end, incredibly subsidized codex/claude plans make it tough to compete for power users.

with freecode, we’re happily embracing it. i realized this was possible after the hard reality of shutting down my company hit: i could no longer charge my many coding plans to the mercury.

on the hunt for a solution, i discovered chinese station relays — massive operations to resell coding agent plans for a ~97% discount. these operations are incredibly sophisticated and involve stealing data, making some margin, and powering massive distilling campaigns (the three meals).

we don’t use these to power freecode, but they made me realize:

• there’s a middle ground between 5% volume discounts and 97% off we can use.
• thanks to this distillation, open-source models have gotten really competitive.
• previously, i was against chinese models because i’m patriotic, but i realized there was nothing more american in the 21st century than moving work to china.

the good ol’ days

it’s been a while since i really enjoyed coding. let’s be honest: unless you’re doing some next-level frontier stuff, it’s solved.

which is kind of a shame, because you don’t get that same dopamine hit when your code finally compiles.

i used to love the problem-solving and spent years teaching others to code to unlock it for them.

working on freecode has restored my love for the art of great software engineering. i have EXTREMELY fond memories of making little games in scratch, hacking away on replit because python didn’t run on my chromebook, and thinking i was so cool because i had the paid version of PyCharm from the GitHub student pack.

i’m so excited for freecode to be the entry point for students and beginners to software :)

design

the one part of swe that i still enjoy is design. it’s not even close to being solved, and i’m learning it at the same time, so it’s super rewarding to finally get it right.

my trailer was originally a compilation of freecode’s design iterations:

i spent weeks tuning the brand/landing page. rightfully so — the original ones were terrible — but i probably should’ve launched a month ago.

i’m hoping that great design sets freecode apart. there is so much slop out there, my competitors don’t seem to care, and people want to use beautiful tools.

what’s next

right now, freecode is a “coding” company, but what’s special is our ability to monetize productive ai usage, not being a coding agent. when the rest of the world catches up to sf in ai usage, i want people to reach for freeWrite/freeDesign/freeAgent first.

there’s a reason the most iconic consumer companies are free.

faq

how are you really affording this?

• we place ads in the terminal and negotiate for incredibly cheap inference. we’re also exploring licensing eligible agent traces and usage data to help fund free access. we’re able to provide opus 5-equivalent models all day to most users.
• if you’re curious about saving money on inference or monetizing with AI, reach out.

are you raising?

i’ve been surprised how this comes up long before any other questions. to be honest, freecode was just to get me off my feet. i thought there was going to be a small window where inference was a little bit cheaper than ad revenue.

happy to say that i’m very wrong, and this turns out to be a massive opportunity. i’m hoping we’ll be profitable very early, and i’m quite frugal (have been living on an air mattress for the last 6 months in the dungeon), but having experienced operators and support sounds incredible.

putting together a pre-seed later this month.

how can i help?

• try freecode! i’m expecting most of the people reading this already spend thousands on inference and want a more powerful tool, but freecode is surprisingly capable, fun, and your feedback is invaluable.
• join the discord :)
• facetime or text me with your feedback on this memo/product.

P.S.

i have to say, this little writing freestyle experiment was great. it’s 4am, and i still need to style the whole page, but freecode is on it ;)