Liam Germain

The man is back

In celebration I'm going to bump some teams off the waiting list for linear.app/diffs, hmu with your best sob story

/main-bus

Come join the best team around!

Our portfolio company @sandstonehq is hiring solutions engineers, CSMs, forward deployed engineers and more. Built by former GCs and legal ops leaders, and backed by Sequoia, SV Angel, and over 20 GCs, Sandstone is the platform for AI-native legal departments. Come work with the amazing team @nifleisher @cc_jarryd! Apply here: sandstone.com/careers

@cryptobyrde amen

They need to provide services to companies that need to handle law, not to the law firms

You guys don’t get it yet. Everyone keeps saying AI is going to replace lawyers. I don’t think people understand how this actually plays out. Let’s say you use AI to draft a contract. The contract misses something important. A year later it costs you two million dollars. What do you do? Right now, you sue your lawyer. In the AI world, you’d sue the AI company. Two things can happen. Option 1: The AI company has liability for legal advice. If that’s the case, every AI company will immediately stop letting consumers use AI for real legal work. The liability risk is massive. Option 2: The AI company has no liability because of disclaimers. If that happens, every state bar in the country will say consumers are being exposed to unregulated legal advice and call it the unauthorized practice of law. And they’ll shut it down that way. Either path leads to the same outcome. Consumer AI will be limited to generic “Wikipedia-style” legal information and LegalZoom level document prep. But the real AI tools? Those will live inside law firms. Lawyers will use them to move faster, analyze more data, and run way more matters at once. The M&A lawyer doing 5 deals at a time will do 50. Trial lawyers will run far more cases simultaneously. The idea that AI replaces lawyers probably dies. The more likely outcome is that AI supercharges the best lawyers and makes the profession even more profitable than ever.

PRDs are more alive than ever. When the cost of implementation goes down, describing what to implement and why is where all the leverage goes.

@sarthakgh In-house >

The product and business model battleground in legal AI is whether to build for law firms or in-house lawyers

I spent the last few days pushing Anthropic's VLIW performance take-home kernel optimization challenge to its limits using Claude Code as an orchestrator. The task: schedule operations for a custom VLIW SIMD architecture running a tree traversal with hashing. 256 items, 16 rounds, 5 execution engines with different slot limits. Starting point: 147,734 cycles (naive) Where Claude Code landed: 1,105 cycles — a 134x speedup The journey involved hundreds of AI agents across several dozen iterations, exploring every angle: hash algebraic merges, L4 tree caching, DAG-based list schedulers, 250K+ parameter configurations, emission order sweeps, store engine exploitation, and loop-based kernels. The last gain was 2 cycles, found by sweeping 103,000 configurations. I then had Claude write a formal lower bound proof in Lean showing the kernel cannot run in fewer than 1,081 cycles — proven from load engine capacity (2,089 ops at 2/cycle = 1,045 minimum) plus unavoidable dependency overhead. For verification I relied on Anthropic's test harness and extended it with randomized parameter testing and also extended it with Kernel Optimization Fun's output index verification. What struck me is how good Claude Code has gotten at orchestrating optimization work. It ran teams of 10 parallel agents in isolated worktrees, each exploring different hypotheses. Agents communicated findings, dead ends propagated instantly, and the system converged on proven optima. The DAG scheduler that broke through a 6-iteration plateau came from Codex (gpt-5.3) running through Claude Code's MCP integration — multi-model orchestration improving the solution. Proebsting's Law says compiler optimizations double program speed every 18 years. AI agents with the right tools are compressing that timeline dramatically — not by improving compilers, but by doing the work compilers can't: reasoning about problem structure, exploring architectural trade-offs, and proving bounds. The 134x speedup here came from algorithmic insight (merged hash stages, path-bits scheduling, bias-free C5), not instruction selection. Wild times for performance engineering.

i knew legal tech would find its way to x for the next generation

the most impressive thing about ramp is not their product, their tech, or their design, but the fact that they got a bunch of smart engineers to think they’re changing the world…with spreadsheets

An American could land on the moon in this freeski big air final and they’d give him a 92

Update, we have figured it out.

@cognition please advise what to do when the team loves the new code review tool but you work with someone also named Devon

@zeddotdev @deanwball i love zed but i have gotten some runaway oxlint memory issues lately, is that just me (at least that is what cc tells me the issue is)

This is true of the Claude app as well. All other aspects of my machine function normally; Claude Code is a resource hog but not machine-stoppingly so. there is something wrong with these GUI apps and it seems not insane to wonder if the vibe-coding helps explain it.

interesting @zeddotdev is becoming a much more enticing option now that i spent less time using AI in my editor. if i'm using @openai codex app, opening zed for small edits is insanely fast. they were playing the long game 👀

@rauchg As well as sandstone.com

You know what supreme.com, tailscale.com, cohere.com, bfl.ai, lumalabs.ai, cerebras.ai, sierra.ai, frame.io, heygen.com, helsing.ai, and 1000s more have in common? They're powered by Sanity and Vercel! Now we're making it even easier to use this dynamic duo ↓

Officially launching X API Pay-Per-Use The core of X developers are indie builders, early stage products, startups, and hobbyists It’s time to open up our X API ecosystem and instill a new wave of next generation X apps We’re so back. developer.x.com

you pity the moth confusing a lamp for the moon, yet here you are confusing a screen for the world