Satyam

I built keys: a CLI tool to manage API keys locally from the terminal. → Store, search, and copy keys instantly → Interactive TUI with fuzzy search → Export to .env in seconds → Profile support for multi-project isolation → Everything stays local on your machine brew install stym06/tap/keys github.com/stym06/keys

@eatonphil uvx

what's everyone using for python dependencies outside of pip and uv?

why are you not typing mail dot google ???

@championswimmer On the other hand, in big tech, AI usage is now factored into performance reviews. If someone’s usage is below the team’s average, it counts against them: probably aimed at coasters.

I'm not gonna say this is a bad thing without looking at the limit first. If it is like $10/day limit, it is stupid. If it is $200/day it is probably okay If someone is hitting $200 everyday, they should have a good explanation, or they are just creating shovels for lulz.

who needs Kafka when you can build a queue on s3 with a single JSON file? Turbopuffer did just that. In their very popular article that was on the front-page of hacker news for a while, they described how they: • used a single queue.json file • all through a single, stateless broker • utilized the so-called "group commit" (fancy word for batching) • added heartbeats to avoid zombies hogging tasks • leveraged S3's compare-and-set operations to ensure consistency And more. The way they handled bootstrapping clients to the broker, all via S3, was really elegant as well. Why didn't they just use Postgres? See my full video here 👇

sam atman it is

@_BetterDiscord_ iterm2 is peak guys..idk why yall switching terminals

kitty terminal doesn't seem so bad after all...

You need to create a CDC pipeline. 1. Set up Debezium to write your database changelogs to Kafka 2. Set up a spark job to consume these events and write to S3 using Apache Iceberg. 3. Set up Trino/Presto for SQL querying these warehouse tables. 4. Set up Apache Superset for visualizing and dashboarding. All open source

Currently we are figuring out how to manage analytics in-house for around 100M events a month. From the outside, analytics looks simple. Track events, store them, query them. But when you actually sit down to build or even evaluate it at scale, reality hits differently. PostHog is there. Good product. But self-hosting it at our scale needs a big machine and serious infra. Paid tools are convenient, but expensive at scale. Some cheaper options exist, but they do not give us the kind of per-user tracking we need to create business funnels. As a bootstrapped company, you cannot just say, “Let’s buy this tool.” You have to ask: - Do we really need everything it offers? - Can we build only the limited parts we actually need? - Can we keep the storage cheaper? - Can we work with a TTL of 1–2 years instead of keeping everything forever? Reading. Discussing. Trying. Failing. Reworking. Learning again. A lot of people think bootstrapped companies move slower because they spend too much time thinking about cost. I think the opposite. - Cost makes you think deeper. - It forces you to understand the system properly. - It pushes you to separate what is essential from what is just nice to have. We are still figuring it out. All of us are learning on the go. But that is also the beauty of building this way. When you cannot throw money at every problem because you don't have enough, you learn to trade-off. Glad that we took a call to build a end to end prep platform, instead of a typical course platform hosted on some third party site.

can claude one-shot clean this?

@championswimmer did you mean anthropomorphing?

The one problem with anthromorphosing "ai agents" like humans is that the quality of output of the AI agent is much more unpredictable than humans. Someone who is sloppy is always sloppy and one who produces diligently checked work, always produces quality work. AI output on the other hand, largely depends on during the task, what all inputs it encountered. Even well framed system prompts can be tripped up by the model simply going through bunch of shitty (or malicious) inputs.

why does every product out there has the same install script? curl xyz | sh install .sh like bro I'm not gonna run any shell file from any website. Just give me brew

@arpit_bhayani @siddontang oh then might as well just use duckdb to write and have an upload thread to upload the db file at the end or at intervals.

@stym06 @siddontang yes. I tried that, but it did not 'completely' serve the use cases that I wanted for mine. I wanted to minimize sync calls and have minimal files in the cloud for one embedded database (to keep cost minimal). I have a write-once use case and that at the end of the session.

Your SSD is a cache. S3 is the real database. That sounds wrong until you see the math: • S3: $0.023/GB/month • EBS gp3: $0.08/GB/month • 3x replication on EBS: $0.24/GB/month At 10TB, that's $230 vs $2,400/month. The trick is making S3 feel like local disk. That's what disaggregated storage engines do — hot path stays fast, cold path stays cheap. The future of databases isn't faster disks. It's smarter caching on infinite storage. The next challenging question is how to build a latency-sensitive OLTP database based on S3. #CloudNative #S3 #ObjectStorage #DatabaseEngineering #DistributedSystems #TiDB #DataInfra

@steipete @shujanshaikh @openclaw same tired of these emails now.

@shujanshaikh @openclaw Yeah there's an insufferable amount of grifters and GitHub makes it too easy to create spam.

someone is sending false emails in the name of @openclaw @steipete

@arpit_bhayani @siddontang there is rocksdb for cloud as well github.com/rockset/rocksd…

@siddontang Totally. I am building S4DB for a use case I needed. It is an embedded database written in Python, with S3 as its storage. The DB is version is semi-functional and currently a work in progress. github.com/arpitbbhayani/…

Dependable always beats the brilliant; if people can not count on you, nothing else matters. In any team or company, reliability is one of the most powerful ways to earn trust, respect, and influence. It does not mean overworking. It means people can trust your actions. If you notice, I did not say "you get things delivered" because that is not the only thing that matters. Here are some simple ways to signal reliability at your workplace - Do what you say you will do - Meet deadlines - Be punctual - Be available and helpful - Run toward problems - Communicate proactively - Follow through and document - Own mistakes quickly - Build a consistent track record Reliability does not always get celebrated, but people do notice it. Over time, reliability compounds into reputation, and reputation opens doors to new opportunities. If you want to build real trust at work, start by being reliable. This does not require doing one big thing but rather a lot of small things every day. The most impactful projects do not go to the loudest or the most brilliant; they go to the ones who can be trusted to deliver. Be the person others can count on.

@TheHackersNews who uses telnet in 2026

⚠️ WARNING - An unpatched critical telnetd bug (CVE-2026-32746) lets attackers gain full system access with no credentials. One connection to port 23 is enough to trigger memory corruption and execute code as root. No patch yet. Prior telnet flaw is already exploited in the wild. 🔗Read → thehackernews.com/2026/03/critic…

Remote work deprives you of those small discussions across the office desk. There’s an uncalled Slack fatigue when you want to ping someone to discuss something. It’s not the same as patting someone on the back and asking for help. There’s also this feeling I miss: Knowledge Osmosis. You learn a lot just by being around people who know their stuff.

@atulit_gaur looks like you watched the video x.com/0xuraraka/stat…

when ilya sutskever joined geoffrey hinton’s lab at the university of toronto, it was a small, almost rebellious place. neural networks weren’t fashionable. most of the field had moved on. hinton hadn’t. ilya stood out, but not in an obvious way. he was quiet, intense, and often thinking in directions that didn’t immediately make sense. there were moments hinton wasn’t fully sure about him - not about his intelligence, but about where it would lead. but hinton had a rare trait as a mentor. he didn’t force direction. he gave space. and ilya used that space differently. while others followed safer paths, he kept pushing on deeper ideas like sequence learning, optimization, representation. things that would later become fundamental. then came alex krizhevsky, and together, in that same lab, they built alexnet. 2012 changed everything. alexnet crushed imagenet. not by a little, but by a gap so large it forced the entire field to rethink. suddenly, the thing hinton had believed in for decades actually worked. that lab - small, stubborn, ignored - became the birthplace of modern deep learning. and ilya was at the center of it.

i wonder when i stopped terminalmaxxing and fancy ohmyzsh commands. now i just ask claude to do it for me.

The feeling of getting stuck, figuring out the fix in your dreams, waking up, and coding it.

@Amank1412 because its not cheap as web or android. first you need a mac and then... ios dev needs you to have the $99 apple developer account and native apps can only be built using xcode on mac.

India’s dev scene is full of > web developers > app developers but barely anyone says “I’m an iOS developer” Either iOS dev is still super niche here or everyone just follows where the users are.