Gary Seiler

35 yrs hardware. Now I ship software from Linux.

Verqury + FlawedWorks.

Still an interstellar explorer.

[Gary Seiler @garypseiler]

35 years I was stuck in hardware. Software arrived on a disk, then in a cloud, and all I could do was live inside someone else's UI.

Those shackles are off.

I ship iOS apps from Linux with no Mac. I built Verqury because every AI coding session lost the plot. Two apps on the App Store this year.

The limiter is no longer "can I build it." It's "which idea is worth the next 48 hours."

If you're a hardware person, an operator, or a Linux holdout who thought software was a priesthood — it isn't anymore.

[Gary Seiler @garypseiler]

Like a synthetic sloth.

[Brett Adcock @adcock_brett]

15 feet up. 15 feet down. Fully autonomous

[Gary Seiler @garypseiler]

Grok- this is solid guidance; can we do these? “Done. In ~/docs/training. What’s next?”

[Ahmad @TheAhmadOsman]

The Ultiamte Step-By-Step LLM Engineering Projects Roadmap

- Build a tokenizer
- Learn embeddings
- Implement RoPE / ALiBi
- Hand-wire attention
- Build MHA
- Build a Transformer block
- Train a mini-former
- Compare objectives
- Build sampling
- Speculative decoding
- KV cache
- MQA / GQA / MLA
- Long context
- FlashAttention
- Hardware budgets
- Toy MoE
- Sparse model trade-offs
- State-space / linear attention
- Diffusion language models
- Data pipelines
- Synthetic data
- Scaling laws
- SFT / DPO / RLHF / GRPO
- Quantization
- Serving stacks
- Eval harnesses
- RAG
- Tool use / agents
- Vision-language adapters
- Interpretability
- Red-team suite
- Full capstone model system

[Gary Seiler @garypseiler]

I’ve been most impressed by the speed and efficiency of Grok. I was building with Claude Code, ran out of tokens, said to Grok Build “look at this repo, the session state, and the build plan”. Grok basically said “I got this” and just picked right up and finished the project.

[Morgan @morganlinton]

I am going to start to share the leaderboard at VulcanBench more regularly.

VulcanBench is the only benchmark I know that benchmarks across effort levels, on an eval suite of 100% real engineering tasks.

Getting ready to sunset eval suite 3 and eval suite 4 is almost ready to go.

Here's the current leaderboard for Eval Suite 3, Grok 4.5 High is the highest scoring model still with Fable 5 Low, yes Low, right behind it.

What I've been able to uncover with VulcanBench is that way too many people are running models at Max effort, thinking that buys them more accuracy, when really it just costs more and uses more tokens so takes more time.

If you're still in the mode of, tell my agent to do something then go get coffee, you're probably still living in the past. You can move faster, with higher accuracy, the key is not thinking you need Max effort all the time.

[Gary Seiler @garypseiler]

Alright- I’m up and ready for all the questions. Go

[Gary Seiler @garypseiler]

Grok Bot did not make me any money today. Find me some coupons or something dang it

[Gary Seiler @garypseiler]

I think this will be so short lived. and the impact of those applications and code so small that it won’t matter. This train is moving so fast. This stuff being built is as transient as GW Basic. Stuff on a 10” floppy that gets tossed in a drawer in 3 months and chucked in the trash when found 10 years later

[Gary Seiler @garypseiler]

I look on here and feel like most of what is being said is just a repeat of what I’ve already read.

[Gary Seiler @garypseiler]

/Morebetterer

[ClaudeDevs @ClaudeDevs]

You can now set Claude Code's output style to Concise.

Claude leads with the result, keeps responses short, and still gives full detail when you ask.

Turn it on in /config → Output style, or set "outputStyle": "Concise" in settings.json.

[Gary Seiler @garypseiler]

@vineerpasam I bet he’s a Sr. Bus Driver though

[Gary Seiler @garypseiler]

They are like “all you want is a damn boat ride, but you’re sitting here telling me how to build a boat!”

[Gary Seiler @garypseiler]

I’ve seen a lot of negative posts about Claude performance lately. I’ve not experienced these issues at all. I think there’s a compulsion to give too much information.

[Gary Seiler @garypseiler]

Working with a smart model is like dealing with a person with ADHD and a 200 IQ. They are distracted and agitated with your slow and primitive thoughts.

[Gary Seiler @garypseiler]

@dhh I’m picking up what you’re putting down.

[Gary Seiler @garypseiler]

@danielkleach Yes please

[Gary Seiler @garypseiler]

Scanning this feed today - the vibe is that AI adoption and tooling is accelerating. I think Grok Bot was the next unlock.

[Gary Seiler @garypseiler]

If your job is easy enough that it can be automated, then really why should a human do it? Humans should be doing human stuff. Machines should do the rest. It’s demeaning to say “just so humans have something to do”. Are we so shallow that we can’t find creative outlet without an hourly exchange of labor for money?

[Gary Seiler @garypseiler]

Brilliant work!

[X Freeze @XFreeze]

This entire SpaceX launch graphic was put together with Grok Build

I wanted a visual showing all 100 SpaceX launches of 2026 so far

Grok Build was able to pull together the exact date, mission and image for every single launch, organize all 100 chronologically, and then build and edit the final graphic

This is the kind of thing that normally takes hours of research, downloading images and manually arranging everything

Grok Build handled the entire workflow by itself and turned hours of work into something dramatically simpler

[Gary Seiler @garypseiler]

Small, unmarked bills, please

[Andrew Yang🧢⬆️🇺🇸 @AndrewYang]

I think an AI dividend is necessary.

[Gary Seiler @garypseiler]

Hopefully a smaller unit, but conceptually correct

[Scott @scottstts]

I predict that in 5-10 years every household will have a local AI server that serves all the AI workload of the house, digital or robotics. It’ll be as common as a WiFi router or a fridge

it creates a private mesh network to link all the devices in the house and does inference completely locally, with built in data vault resistant to remote data hacking or physical tampering. That way you can have every single piece of information about you and your family stored in there and never leaves it, and AI can use them to provide most accurate and customized services