Ketan

Software engineer by day, indie developer by night.

Building

[Ketan @_ketansahu]

Happy New Year to all the builders out there 🎉

I really want to build a SaaS this year and quit my job.

[Hari Prasad @hariprasad_bg]

This bro said he didn't clone the @kapilbuild app.

Even I didn't.

Launching https://t.co/mBF1Vwcf6f (completly free) It has 9 views and a dark theme.

Some screens better than https://t.co/LyosKwnm2U

Mac storage visualizers aren't new for sure 😜

[Ketan @_ketansahu]

https://t.co/jbXjwTTAWc

[Ketan @_ketansahu]

OpenAI, Anthropic and Google have been quietly meeting since July to build an AI safety standards body. The "slow down" statements this weekend were not the start of the conversation. They were the public face of a plan that's been running for months.

The plan is shared testing before release. Independent evals, pre-release safety reviews, standard risk assessments. The three biggest labs building the referee for their own race.

The wild part is how differently they answer the same question. Amodei wants government in the room. Altman says if the government won't back it, the labs build the body themselves. Hassabis proposed a FINRA-style standards body back in July. Three companies, three answers, one table.

For the rest of us building on these models: the rules of the game are being written right now, and the people writing them are the ones playing. Watch where this lands.

[Ketan @_ketansahu]

@MohammedDa14859 Which subs do you buy? $20?

[Ketan @_ketansahu]

Today, Anthropic "raised" Claude Code weekly limits by 25%. If you've been paying attention, you just lost 17%.

Everyone's been running on a temporary 50% boost since May. 150% of the old standard. That promo just ended. The new permanent number: 125%. The announcement calls it a raise. Compared to what you had this morning, it's a sixth less. Anthropic's own follow-up confirmed it, after the first announcement got so much heat they deleted it and reposted with the clarification.

The pattern is worth naming. "25% more than baseline" and "17% less than you have today" are the same sentence, dressed differently. Nobody emails you to say your quota dropped. They announce an improvement.

For anyone building on subscription AI: read the denominator. And budget your week like the meter shrank, because it did.

[Sawyer Merritt @SawyerMerritt]

Very impressive FSD maneuvering on the edge of a cliff road in Slovenia.

Tesla FSD (Supervised) was approved in Slovenia last week.

[Ketan @_ketansahu]

Shipped the AI summary feature in my extension.

Chat with video and more features are coming.

[Ketan @_ketansahu]

The AI feature I've been promising for weeks is finally out. In beta, but out.

It was supposed to be the easy part. My extension already saves your timestamps, your notes, your tags. A summary on top felt like a natural next step. Spoiler: nothing about it was easy. I rewrote the flow twice, broke my own dashboard once, and lost a whole weekend to a bug that turned out to be one wrong header. The usual.

But it's live now. Save a video, and the AI writes a summary for it. Free users get 3 a month, Pro gets 30. I kept the limits low on purpose, it's a beta, and I'd rather under-promise than have it blow up in someone's face.

Honest status: still zero subscribers, still zero dollars. But the product finally does the thing I described to people when I started. The first subscriber comes when it comes. Until then I'm just glad people can finally try the feature I kept talking about :)

https://t.co/EW0h9oD4MY

[Ketan @_ketansahu]

There's a website that predicts when Codex resets your weekly usage limit by watching Tibo's posts.

41 verified resets tracked, a 24-48h forecast, and even a personal timer that counts down your own window.

Someone built this because waiting for a reset feels like waiting for rain in the desert.

https://t.co/5I43ymOx1p

[Ketan @_ketansahu]

I'm mind-blown by how good Astra, it did all of the tasks in two prompts that I have been procrastinating on for almost two and a half weeks now 🤯

[Ketan @_ketansahu]

@reach_vb Amazing! Really thankful :)

[Ketan @_ketansahu]

I was running a Hermes agent and multiple Grok bots, and god, managing so many agents at the speed they work is not good for focus. I was always jumping from one task to another, and that was ruining my focus.

Have stopped using Grok bots for a while now and only using a single agent.

[Ketan @_ketansahu]

I finally shipped the thing I kept teasing.

AI summaries are live in beta on https://t.co/BI5DMuxmE0. Log into your dashboard, click any saved timestamp, and it writes you a summary of that video. The part I like most: it also suggests the moments worth saving. So you don't have to rewatch a 40-minute video to find the one useful bit.

Free accounts get 3 summaries. Pro gets 30 a month.

Built the whole feature with Codex, which is a whole other story.

Link 👇

https://t.co/EW0h9oDCCw

[Ketan @_ketansahu]

The most effective marketing I've done on this app isn't posting. It's replying.

Every evening I go looking for small builders who posted their day-3 update or their first MVP and got nothing. Not the big accounts, those comment sections are a party I wasn't invited to. The tiny posts. Five likes, zero comments.

Here's the thing nobody tells you. When a post has 200 comments, nobody reads yours. When a post has zero, the person who wrote it reads every single one. They see your name, they click your profile, sometimes they follow.

I did this for a week and the numbers moved. 2.9K impressions, up 181%. Profile visits up 233%. Eight new followers in two days, all from comments on posts nobody else bothered with. My own posts never did that.

The value of a comment is inversely proportional to how many comments the post already has. That's the whole trick.

Anyway, back to waiting for my first subscriber :)

[Ketan @_ketansahu]

Switched to Codex on Astra recently and honestly, I feel useless now.

It just keeps working. For hours. I went to the gym, came back, and it had rebuilt the AI summaries beta for https://t.co/BI5DMuxmE0 with a better design than the one I had in my head.

I'm the project manager now. I'm not sure the project needs one.

[Ketan @_ketansahu]

Chronic lower back pain, baseline 3-4/10 every single day. So I built my whole gym plan around not making it worse: no kettlebells, no heavy deadlifts, every lift has a back-safe alternative.

This week I hit 220kg on the leg press, 12 clean reps. And for the first time in months, the pain is showing up less. Same intensity, but I'm getting confident with my back now, and confidence changes everything.

The plan is working. The numbers are going up, and the pain isn't following.

[Ketan @_ketansahu]

DeepSeek released their smallest model today. It was so good they're retiring their flagship.

V4.1 Flash is the first model in their new architecture family, and it beats V4 Pro on performance, cost and speed. From Monday, every call to their pro model gets routed to the flash model and billed at flash prices. They retired their flagship because the cheap one beat it.

It's also DeepSeek's first model with native vision. Text-only DeepSeek is over. 1M context, and they cut API prices again on top of it.

For anyone building products on these APIs, this is the part to watch. The model that wins isn't the smartest one. It's the one fast and cheap enough to live inside a real product. Speed and price beat benchmarks, every time.

Give me a slightly worse model that responds in half a second over a genius that makes my users wait. Every single time.

https://t.co/TbfTUjDj79

[Ketan @_ketansahu]

Working on my sleep cycle these days.

If you have not tried yoga nidra yet, do give it a try it helps you fall asleep faster.

[Ketan @_ketansahu]

On Codex $100 now. Testing how far and how fast I can work with Astra :)

[Ketan @_ketansahu]

Which one should I go for?
Codex $100 plan or Claude $100 plan?

This is the first time I would be buying the $100 plan, so I really need your help in finding out which one is best for a solo developer. I code daily.

[Ketan @_ketansahu]

DeepSeek released their smallest model today. It was so good they're retiring their flagship.

V4.1 Flash is the first model in their new architecture family, and it beats V4 Pro on performance, cost and speed. From Monday, every call to their pro model gets routed to the flash model and billed at flash prices. They retired their flagship because the cheap one beat it.

It's also DeepSeek's first model with native vision. Text-only DeepSeek is over. 1M context, and they cut API prices again on top of it.

For anyone building products on these APIs, this is the part to watch. The model that wins isn't the smartest one. It's the one fast and cheap enough to live inside a real product. Speed and price beat benchmarks, every time.

Give me a slightly worse model that responds in half a second over a genius that makes my users wait. Every single time.

https://t.co/TbfTUjDj79