pranjal (@betterclever) header
pranjal (@betterclever) profile picture

professional claude yeller. currently cooking @makeo_fun

J space
Joined December 2015
Photo by pranjal (@betterclever)Photo by pranjal (@betterclever)Photo by pranjal (@betterclever)Photo by pranjal (@betterclever)Photo by pranjal (@betterclever)Photo by pranjal (@betterclever)Photo by pranjal (@betterclever)Photo by pranjal (@betterclever)Photo by pranjal (@betterclever)Photo by pranjal (@betterclever)
Meet Wraith 👻 the fastest terminal on the web. 👻 Ghostty's terminal engine ⚡ Rust WebGPU renderer, off the main thread 🖼️ Kitty graphics: real images in the terminal 🔤 Ligatures + Arabic & Hindi shaping 🔌 Drop-in xterm.js API, one import 🎮 DOOM in your browser (in the terminal): 400+ FPS
15
13
6
283
17,100
vulnerabilities at this scale are out in the wild! imagine the scale of the things that are about to come.
Our security researchers built an autonomous cyber vulnerability discovery, exploitation, and patching pipeline. In a single run with an upcoming cyber model, it validated and generated patches for 948 local privilege escalation vulnerabilities in the Linux kernel at a total API cost of $1.2 million.
2
119
I swear there hasn’t been a single post of OpenCode people on my timeline doing normal stuff.
i hate this team bro
2
13
682
I was not aware of the incompetencies of CERT-IN to this degree. @ni5arga told about them in his session at @web3privacy congress and it was baffling to hear about the state of preparedness that we have as a country in the current state of the world. honestly people like him are doing a real service to the nation.
CERT-IN still requires researchers submit PDF forms in order to report security vulnerabilities in the big year of 2026. Is this the agency that is expected to protect Indian computer systems from swarms of malicious AI agents and non-stop exploitation attempts?
4
267
a feature of @EFDevcon being in India is that all the racist fucks are already out. good riddance honestly. the world is better without you.
1
6
682
This is the optimistic OpenAI I want out there. The model absolutely cooks by the looks of this video generation. Can’t wait.
There are many futures. Which do you want to build?
1
175
we are changing the rules. the next mod drops while you're still playing 🎮 type what you want, mid-match: "give me a nuke" 💥 "make the ring ice" 🧊 "knights jump twice" ♞ the game rewrites itself live. no download, no restart. real-time AI modding is coming to @makeo_fun ⚡
Built an in-game AI feature for @makeo_fun. Mid-match, type what you want and the AI rewrites the rules of the game you’re playing, live, without stopping the match. Ask for anything: “make d4 and e5 lava” in chess, “give me a nuke” in a shooter. Works in any game built on Makeo. Coming Soon! Inspired by @jarrodwatts’ work on Dream.
1
5
379
this is the coolest thing I saw this week! stochastic parrot crowd in shambles.
1/4 Meet Qwenputer: Qwen3.5-4B with a full RISC-V CPU inside its weights, using the model's own Gated DeltaNet layers as RAM. It runs C and MicroPython in its forward passes, no external runtime. On DeepMind math: base 55–90%, base + Python tool 95–100%, Qwenputer 97.5–100%.
1
1
289
In India, Jio costs way less than this to a user who actually gets the benefit. this law is rent seeking on another level. EU law makers smoke crack.
‼️ French lawmakers voted for an internet traffic tax that the makers of VLC say would cost them more than €1M a year. Content providers would pay €0.055 for every gigabyte sent to French users, as counted by ISPs. Its Socialist authors expect €10B a year. The government opposed it, warning that about half the bill would fall on smaller providers, many of them French. It passed 11 votes to 10, with 35 of 577 deputies present.
148
this is getting too scary but nothing surprising. an year and half back when I started working on shipsec.ai, it was getting painfully obvious that the world was heading here. what was hard was to figure out what's the right shape of the product that would help the users. surprsingly to some of our design partners, problems like "offboarding" the users was way more important than infra hardening, finding and patching exploits in realtime etc. other discussions were just about having enough security for compliance which is a whole theatre at best. no one really cared then. we couldn't really find the PMF then because as we wasted time in talking to the "leaders" of the companies for the products, the right product was already evolving by the security first folks who only focused on the threat ahead and not what enterprises would pay for. enterprises would now happily pay for that. in AI, you ALWAYS need to think 6 months ahead. it might not be obvious to everyone now, but very clear in your head that in 6 months from now, this will be the new normal.
man I have never seen anything like this hacks after hacks after hacks an actual world war is happening on the internet atm
7
344
"with AI there's no moat, everyone can build X in 2 days" well, but are you building it though or crying about the moat yet? AI democratizes capabilities. build first, cry about moat later. through the building process you'll find totally new problems that are not solved by anyone yet.
3
8
340
"fastest terminal on the web" for a terminal under 3 weeks old, written entirely by claude, is a big claim. so i tried to kill it. 🔪 one tab, more terminals until they die 💀 xterm.js at 32. wterm 64. restty 256. zem 384. ghostty-web 512. 👻 wraith: 1,024 at 116 fps. even gets 90+ fps with shaders on 🔥
2
8
488
I don’t understand how the secops can be so bad that you don’t need a multisig or hardware-backed keys for this. DevOps team that are not super paranoid about each action they do are a liability. Always ask yourself as an engineer with access to critical infrastructure that if someone wrench attacked me (the highest tier of compromise imo), would they get the access to critical amount of funds and even in that case the answer should be no. It’s for your own safety. If you provably don’t have access to critical infrastructure and funds in a weak way, the chances of you being personally attacked reduces massively.
Binance's initial investigation suggests that the security incident may involve a compromised DevOps engineer’s laptop, which allowed an attacker to gain access to a service container and submit unauthorized withdrawal transactions. We have found no evidence of any wallet private key leak. We will share a detailed technical post-mortem once the investigation is complete and the facts are fully verified. The estimated total impact sits at around $6.3B. We have worked with blockchain partners to freeze affected assets, with ~$3 billion frozen on-chain so far. Binance will cover all losses from the SAFU fund and the company’s funds. User funds are safe. We will also publish an updated Proof of Reserves once the remediation is complete. As an immediate precaution, Binance has: - temporarily suspended deposits and withdrawals while security checks are completed; - revoked the affected address authorisation; - engaged law enforcement and blockchain intelligence partners to trace and recover funds. Our teams are working around the clock to resolve this issue safely and thoroughly. We are working closely with our law enforcement partners to confirm the identity of these criminals and bring them to justice. Please rely only on official Binance channels for updates. Please feel free to contact Binance Customer Support if you have any further questions. binance.com/en/support/annou…
3
327
Halo 2 update on Zebend, and a correction. I said our Bend verifier was 2.1× faster than the Rust crates. True vs upstream, but I hadn't benchmarked Zakura's crates, which were 4.6× faster than us. Thanks @zkDragon for flagging it. One optimisation pass later: around 2x faster. Zakura still leads, 2.6× tbh I'm just copying the best ideas from Zakura
Privacy on @zcash runs on math. Math is written as code. Code can be wrong. So we rewrote it, with proofs. Meet Zebend: an independent Zcash full node, built from scratch, every consensus check in Bend. Live on mainnet. 🧵
1
2
15
648
Same 392,806 mainnet Orchard bundles, every verdict still matches the official code. Next: a native 64-bit field multiply in our Bend fork, written from scratch, not a link to Zakura's assembly. Numbers and method: github.com/betterclever/zebe…
52
Today we're launching f.xyz. Tell F what you think will happen in the world. It builds a portfolio around it, shows how it would have done, and runs it for you. Tag @fxyzai under any post and F will build it right there.
21
19
9
144
13,247
writing zebend has really made me appreciate the efforts of the teams behind @ZakuraZcash a whole lot more i.e. @TachyonZcash and @ValarGroup so many assembly level clever optimizations to get the absolute highest performance. I'm learning so much as claude tells each to me and I figure out how to port that to bend.
Privacy on @zcash runs on math. Math is written as code. Code can be wrong. So we rewrote it, with proofs. Meet Zebend: an independent Zcash full node, built from scratch, every consensus check in Bend. Live on mainnet. 🧵
1
1
10
469
things are getting exciting at @makeo_fun we are getting to a point where we'd be a class apart from any prompt-to-game or social-gaming platform out there and the gap won't be close. onchain games never took off because they were early. we are going to fix it.
2
5
201
I started working on exactly this in July when I dropped this because the models were not there yet. Now they are and this looks incredible! I believe I can take a second shot at this too now.
Today, we're shipping real-time graphics. Using Sol 6.1 Ultrafast we can generate custom overlays, diagrams, and even interactive 3D scenes right on a user's screen, no chat. It's a new visual interface for AI. Here's me learning about LLMs + hardware. Demo:
1
1
10
830