Sponsored by
The World's Biggest Dev Event Hits Silicon Valley
One trip, and your team stops guessing on AI adoption, platform, security, and tooling.
WeAreDevelopers World Congress — San José, CA, September 23–25, 2026, in the heart of Silicon Valley. 500+ speakers and 10,000+ developers working through what actually ships in production: coding agents, evals, MCP, observability, agent security.
Kelsey Hightower. Thomas Dohmke (fmr. CEO, GitHub). Christine Yen (CEO, Honeycomb). Mathias Biilmann (CEO, Netlify). Olivier Pomel (CEO, Datadog). The people actually building the tools you use every day — all on one stage.
Bring a shortlist of decisions; leave with answers and contacts you can still reach after the event. Prices rise as we get closer.
Use code GITPUSH26 for 10% off.
See the whole platform. No guided tour.
Skip the sales call. Walk through Gladly's interface yourself — the AI suggestions, the unified customer view, the full conversation thread. 15 minutes, no installation, no commitment.
Welcome to Grind Engineer, your guide to becoming a better software engineer! No fluff. Pure engineering insights.
📱 Big Tech & AI
OpenAI Launches GPT 5.6 Sol, Terra, and Luna on Apps and API (5 minute read)
OpenAI's three tier model family is now publicly available. Sol is the flagship with multi agent coordination, Terra handles everyday work, and Luna is built for speed and cost. Sol's "ultra mode" spawns subagents autonomously for complex tasks. This is the first time a major lab has shipped three models simultaneously across all platforms.
Scortier's Take: The model tier war is officially on. OpenAI has Sol/Terra/Luna, Anthropic has Opus/Sonnet/Haiku, and now SpaceXAI has Grok 4.5. If you're still hardcoding a single model into your app, you're leaving performance and money on the table. Route by task complexity or get left behind.
Meta Starts Charging for AI with Muse Spark 1.1 (4 minute read)
Meta launched Muse Spark 1.1 with a freemium developer tier and a paid API priced roughly 25% cheaper than competitors. It includes tool use, coding, and multimodal reasoning. Meta's play is clear: undercut everyone on price while leveraging billions in existing infrastructure. The free tier for open source models is over.
SpaceXAI Releases Grok 4.5, an "Opus Class" Model (4 minute read)
Grok 4.5 claims twice the token efficiency of competitors at $2 input / $6 output per million tokens. Musk calls it "Opus class." Whether that holds up in production remains to be seen, but the pricing alone makes it worth benchmarking against whatever you're running today.
Apple Exploring PrismML to Run 27B Parameter Models on iPhones (3 minute read)
Apple is working with startup PrismML to compress Alibaba's 27 billion parameter Qwen model for on device deployment on iPhone Pro. If this works, it means frontier class inference without sending data to any server. Privacy and latency wins in one move.
🚀 Engineering & System Design
Rewriting Bun in Rust: 11 Days, One Engineer, AI Assisted (12 minute read)
One Anthropic engineer rewrote Bun's entire runtime from Zig to Rust in 11 days using Claude Fable 5. The same project would've taken three engineers roughly a year by hand. They fixed 16,000 compiler errors, drove test failures from 972 to 23 in two days, and memory leaks dropped from 3 MB per build call to flat at 600 MB. Binary size shrank 20%. This is the most concrete "AI accelerates real engineering" story I've seen.
How GitHub Gave Every Repository a Durable Owner (6 minute read)
GitHub assigned clear ownership to all 14,000+ internal repositories in 45 days. Active projects got validated owners, abandoned ones got archived. If your company has hundreds of repos and nobody knows who owns what, this is the playbook for fixing it fast.
Cloudflare built Meerkat, an experimental consensus protocol that manages control plane state across their global network without a single central coordinator. For anyone building distributed systems that need to agree on state across regions, this is worth studying.
💻 Dev Tools & Programming
TypeScript 7.0: 8x to 12x Faster Builds (5 minute read)
TypeScript 7 ships with native code speed and shared memory multithreading. Full builds are 8x to 12x faster. If your CI pipeline spends minutes on type checking, this single upgrade could save you real time and money on every push.
Cursor Stats: Top Users Write 40,000 Lines Per Week (7 minute read)
Pragmatic Engineer's analysis of Cursor data reveals wild productivity gaps. Top users generate 40,000 lines weekly versus median users at 700 lines. Input tokens make up 90% of usage, meaning the best users are feeding context, not just accepting completions. The skill gap in AI assisted coding is enormous.
Martin Fowler on Local Models for Coding (8 minute read)
Martin Fowler's practical evaluation of running LLMs locally instead of cloud APIs. Covers model selection, real performance comparisons, and when local makes sense versus when it doesn't. If you want AI coding without sending proprietary code to external servers, this is the most balanced take available.
🎁 Miscellaneous
A Brown professor suspected AI cheating and switched to in person finals. Average scores dropped from 96 to 48. Twenty seven students dropped or skipped the exam entirely. This is the clearest evidence yet that some students are offloading thinking entirely to AI, and it raises real questions about what "learning" means in 2026.
Lenny's survey data shows the tech workforce diverging into two camps: those who've embraced AI tools and feel more productive, and those who feel increasingly replaceable. Career trajectories are splitting based on AI adoption speed, not years of experience.
Former GitHub CEO Launches Entire: Git Hosting for AI Agents (4 minute read)
Nat Friedman's new company Entire is a Git hosting platform designed for AI agents to read, write, and collaborate on code natively. It's built for the assumption that most commits will soon come from agents, not humans.
⚡ Quick Links
OpenAI's voice mode now spins off complex tasks to GPT 5.5 mid conversation and holds sessions over an hour.
Bezos contributed $2B alongside external investors, targeting New Glenn return to flight by year end.
Major supply chain security improvement: automated script execution now off by default in new npm.
Security researchers demonstrated a vulnerability in GitHub Copilot that exposed private repository contents.
Deep dive into HubSpot's distributed vector database architecture at massive scale.
Gamma CEO shares how they reached $100M revenue with a team of 50, plus the critical mistakes along the way.
Uses open source uBlockOrigin filter lists to strip video ads including YouTube pre rolls.
Workers can now have a dedicated cache in front of them, major performance upgrade for edge compute.
See you in the next one!
Scortier, Signing Off!