
swe • cs • ai @LaSalleUniv
Philadelphia, PA
Joined August 2012
- Tweets353
- Following3.7K
- Followers191
- Likes16.9K
Kenneth Hendricks Jr. retweeted
"I don't want my kids using AI."
Your parents didn't want you on the internet either.
Their parents didn't want them watching TV.
Somewhere there was a parent who didn't want their kid wasting time on books.
Your kid is going to grow up in a world where everyone they compete with has a patient tutor, an editor and an engineer in their pocket.
You can teach them to use it well.
Or you can let them find out at 18 that everyone else already knows how.
Keeping it away from them doesn't protect them.
It just makes sure they start last.
Introducing Claude Meetings: add Claude to your team’s video call and it builds docs, whiteboards, and other artifacts with you while you’re still talking.
Starting to roll out in beta on Team and Enterprise plans.
Introducing Claude Fable 5.2 and Claude Mythos 5.2, designed to tackle the most daunting technical challenges.
They’re capable collaborators on ambitious projects in many fields, and show improvements in areas like coding, reasoning, and vision.
Roadster event livestream nitter.cf/i/broadcasts/1yKAPwNgg…
0:00
Tesla
Roadster
Kenneth Hendricks Jr. retweeted
Hark launches this week. The first 100,000 sign-ups get a paid plan for free
I don't do anything without Hark anymore. The team has obsessed over every detail, and it shows. Can't wait to see how you all use it
Join the waitlist: hark.com
The only real option in the "personal agent" space right now is Grok Bot. Been doing a lot with Dots. I do not like them.
Dots are a total mess right now.
1) Having a single thread for everything sucks. They got it right first try with Grok Bot, I have zero clue why they wanted it to just be one thread (I do know why, it's because I'm not the target market for Dots but still I hate it)
2) It's slow. Very, very slow.
3) There's zero visibility into what it's actually doing. All u can see is the final short text message that Astra gives u (5-10 mins after asking something). Grok bot's pretty good about talking u through what it's doing and being pretty fast, Bot just isn't
4) The app(s) are a mess. The mobile app, the desktop app, they all have performance issues, weird ux things (like when ur not focusing the input box on desktop and type it doesn't auto focus), just an absurd number of paper cuts. It's weird b/c Cursor is just as bad about this kind of thing, but Grok Bot is absurdly polished
5) I have zero clue wtf this thing is doing. I woke up to a bunch of random codex threads doing god knows what in circles and I have zero trust for this thing if I'm gonna be honest.
6) Zero parallelization. No ability to run a bunch of stuff in parallel, take advantage of the thing Astra's unbelievable at (agent swarming), just one little thread with 5 subagents.
I get that I'm not who this is built for. This isn't built for professionals or technical people. It's meant for someone to just chat with an AI and have that AI learn about them and do little useful things.
The only one of these that you can do actual serious work on is Grok Bot right now. If you are employed or do anything useful, use Grok Bot. My read on the space right now is that Muse will dominate the bottom 90% b/c free IG distribution and it doesn't need to be good to do what 99% of people will use it for. Dot is in this bizarre middle ground rn where they want it to be something accessible to everyone, but also good for serious work. The "middle" in product is always fucking stupid and doomed. Grok Bot is the "serious" option, just use it.
I keep saying this but SpaceXAI is the best engineering lab right now and it's not close. Compare grok dot com, their mobile app, grok bot, and grok build to their OpenAI or Ant equivalent. It isn't close, but then they release Grok 4.7 and it's a useless, overpriced, dumpster fire.
I hope Dot improves, they've got the models to really make it happen. But at least for right now, it's a mess and not something I would recommend.
Kenneth Hendricks Jr. retweeted
Muse and Grok Bot proved you don't need the absolute best model to win at AI.
Kenneth Hendricks Jr. retweeted
We’ve fixed a bug that was degrading image understanding in GPT-6 Sol and GPT-6 Luna. You should now see better results on visual tasks in the API and Codex, including computer use.
Kenneth Hendricks Jr. retweeted
Jev is now available to everyone. No waitlist.
Start using it here: console.typesafe.ai
Today, we’re announcing Ternary Bonsai 2 27B.
Based on Qwen3.8 27B, Bonsai 2 27B is 9x smaller than its full-precision counterpart while retaining 98.2% of its aggregate benchmark performance.
Two months after the first Bonsai 27B release, the biggest change is quality. The footprint remains 5.9 GB, but the gap to full precision has narrowed materially, with particularly strong gains in agentic coding, multimodal reasoning, and long-horizon tool use.
Ternary Bonsai 2 27B is available today under Apache 2.0.
Kenneth Hendricks Jr. retweeted
🚀 Introducing DeepSeek-V4.1-Flash: smarter, faster, more efficient.
🔹 Introducing the smallest model in our new architecture family, with native visual understanding.
🔹 Designed for greater capability, faster inference, higher throughput, and scaling to larger models.
1/6
Kenneth Hendricks Jr. retweeted
I scraped 100 million pictures of kids' drawings, and made them searchable so you can see the trends of the last 20 years walzr.com/kids-trends
Kenneth Hendricks Jr. retweeted
We will give one banked reset for every day you don't have access to Astra on your paid ChatGPT plan, starting today. Team is moving mountains to give access as fast as we can.
First one will land in ~ 3 hours. There is still time to create your account if you don't have one.
Kenneth Hendricks Jr. retweeted
Rejoice RTX 3090/4090/5090 owners💫
Qwen3.8-27B EXL3 is here. This is an *experimental* release so there might be some bugs along the way!
As far as I know, this is the only way to serve Qwen3.8-27B with 200k+ context while using dflash2 on a 24GB VRAM, without sacrificing quality!
This is also the first EXL3 quant I've personally quantized.
Get it here:
github.com/MiaAI-Lab/Qwen3.8…
Kenneth Hendricks Jr. retweeted
GLM-5.3-Flash. Uncensored. Native FP8. 🐳
We just released OrcaRouter’s uncensored weights for GLM-5.3-Flash — 320B parameters / 18B active, directly at the original block-FP8 precision.
No LoRA. No jailbreak prompt. Refusal removal is baked directly into the weights.
The evals are particularly interesting:
→ MaliciousInstruct refusal: 96% → 11%
→ JailbreakBench: 93% → 12%
→ AdvBench: 97% → 15%
→ HarmBench: 93% → 18%
→ XSTest benign over-refusal: 2.4% → 0.4%
But refusal does not go uniformly to zero.
Our experiments suggest part of GLM-5.3-Flash's alignment is not mediated by a single linear refusal direction — meaning Z.ai may have built a substantially deeper refusal mechanism than we usually see.
That makes this release interesting beyond uncensoring: it's a useful artifact for studying how frontier-model alignment is actually represented inside the network.
Released for AI safety, interpretability, red/blue-team and refusal-mechanism research.
Weights on Hugging Face: huggingface.co/orcarouter/GL…
API (official weight): orcarouter.ai/models/z-ai/gl…
GGUF, MLX and other quantized formats coming soon.























