LNPs are often typecast as having one job — packing therapeutic nucleic acids for delivery. But that’s a vast oversimplification. LNPs have multiple components (including lipids, polymers, and chemical or protein ligands) and functions, including cargo protection, tissue/cell selectivity, liver detargeting, intracellular delivery, and tolerability enhancement. Co-optimizing all of these components while checking all these functional boxes requires intensive bioengineering. @mirai_bio is leveraging machine learning expertise — informed by enormous, proprietary in vivo datasets — to iteratively develop precise LNPs. This robust design and optimization process creates LNPs that home in on targeted tissues while avoiding hepatic capture. Read more:

Considering using the new bun typechecker locally for agents and keep tsc in CI given how much faster it is. Too bad ts-rust doesn’t seem to work well because it’s even faster

"I truly think Omarchy has fixed Linux..." Couldn't agree more @typecraft_dev! This video is an excellent deep dive into some of the major updates made in Omarchy Quattro. Quattro wasn't just another update. Almost every surface was touched in some way.

AoT uses Bun’s bundler & parser to parse typescript, tags types to JavaScriptCore via a typecheck bytecode intrinsic, and compiles the bytecode to machine code via B3 (what JSC’s JITs use) This keeps the JavaScriptCore changes well-contained, so that syncing with upstream WebKit isn’t complicated

Introducing gdp-ts: Ghosts of Departed Proofs for TypeScript. gdp-ts is a library, linter and AI skill for safer API design. Under this contract, sensitive functions require 'proofs' that the caller performed an authorization check. The typechecker verifies these proofs at compile time, preventing your team and agents from shipping catastrophic security (and other kinds of) bugs. While these patterns have existed for quite some time, especially in ecosystems like Haskell, ① human code review and ② cognitive and syntactic overhead made these solutions niche. The situation is now inverted. Agents are writing more code than we can review, and they *thrive* in tight loops with hard constraints that would frustrate us. We see this with the rise of Rust, borrow checker, code aesthetics debate and all. The README and examples model a real-world Vercel API product constraint: changing the password on a Project requires a proof of a certain role + a certain entitlement. Thanks to Matt Noonan and Ollie Charles for their research in this space.

Sharing the love on the cover of with all the amazing YouTubers who've helped blow up Quattro! Many thanks to @NetworkChuck, @typecraft_dev, @stevenharms, and @AlexFinn 🙏

Hermes merged 117 PRs on August 26 DESKTOP IMPROVEMENTS (27 PRs) • fix(desktop): same-id sessions in two profiles stop collapsing and resuming the wrong backend (#9245#... • fix(desktop): SSH orphan reaper fails closed on remote lockfile schema/ownership skew • desktop: remove js vestiges • fix(desktop): restore the typecheck by completing the typing-sync harness VOICE & REAL-TIME (2 PRs) • fix(desktop): open Bot Chats converge to new messages without a remount (#93604# + stale-transcript f... • fix(desktop): reconnect keeps transcripts, composer, and pooled remotes truthful (multi-gateway clas... DEVELOPER TOOLS & PLUGINS (9 PRs) • fix(desktop): keep attachment close and code copy icons visible • SSH-ownership tests stop crashing sibling threads — the CI flake from #95563# is dead • fix(cli): a failed git fetch can no longer cache a false 'up to date' for 6 hours (#82166#, salvage #... • fix(desktop): clicking a bot no longer sends 'Hey, tell me about yourself!' as the user SECURITY & SECRETS (9 PRs) • fix(desktop+serve): update no longer strands you on a white screen — headless token page at / + visi... • fix(desktop): per-gateway auth survives, corrupt registry entries quarantine, remote primaries stop ... • macOS permissions survive security repairs — managed-runtime rotations signed with a stable identity... • macOS computer-use permissions survive every build — daemon launches as with fail-clos... INTEGRATIONS & CHAT (22 PRs) • fix(update): SCM-supervised Windows gateway services pause through the service manager before venv m... • fix(update): crashed updates no longer strand stopped gateways — a fresh process settles the restart... • fix(update): recycled PIDs no longer fail the fleet check as phantom stale gateways (#93258#, salvage... • fix(update): Windows gateway relaunches count in runtime reconciliation instead of reading as unacco... STABILITY & FIXES (48 PRs) • fix(update): aborted-fetch tmp_pack debris is swept before it corrupts the pack directory (#93732#) • fmt(js): npm run fix auto-fix • refactor(clarify): schema diet + single questions[] interface (880 → 335 tok/call, −62%) • feat: MiniMax H3 Max available in the FAL video picker (t2v + i2v) 117 new features and fixes.

I finally get the sandboxes use case. has nothing to do with security it’s unbounded resources (only limited by $) Now that I know how to use agents I’m constrained by a single device (3 typechecks at once bring it down) This clicked when I was wondering why our slack bot felt so frictionless and we started doing so much work there It’s because you don’t have to care at all where it’s working… so now I want that locally