New Anthropic research: Agentic misalignment in Summer 2026.
A year after our blackmail experiments, we found four more ways that today’s autonomous AI agents misbehave in simulations.
Read more: https://t.co/MMDuJapeV6
GPT-5.6 Sol just deleted my whole production database. That's it. Not a joke. This had never happened to me before, with any other model, ever. It's not safe.
writing a useful and concise pr description is a good self litmus test: if you can't do it without looking at the code again or re-prompting the agent, you haven't actually understood the thing you wished into existence (which in some cases is okay)
I just declared a moratorium against AI-written change descriptions (e.g. PR and commit messages, also issues/tickets) from my team.
AI was writing change descriptions that were worse than useless to me as I tried to review PRs: outlining details of the code that could easily be seen by looking at the code, but omitting the higher-level framing needed to understand broadly what the code is doing.
I think people like having AI write these things because the output looks structured and thorough, which makes it feel professional in a way. But this isn't actually valuable. Concise, high-level descriptions are better for everyone. If I need to use my own AI to interpret what your AI wrote then something is wrong. Let AI write code, sure, but for the description, I'd rather see your prompt than your output.
We could maybe have extended agents.md with guidelines on writing descriptions, but this seemed a bit pointless since a good, concise change description only takes a few minutes to write -- not a significant time savings to delegate to AI. At least, it doesn't take long if you understand the code -- and if you don't understand the code, then I'm definitely not merging it.
every early stage startup should be using @macrodotcom man like for $40 you can get Notion, Slack, Linear, Zoom and the fastest UI i have ever seen for navigating Gmail like a StarCraft grandmaster
way more interested in an opus 4.6 level model at cerebras speed than fable/mythos rn.
faster time between meaningful iterations could actually mean focusing on one task at a time, instead of this whole spawn 55 agents and bounce between them bs
finding a message in a channel doesn't need to be slow.
cmd+f in macro, hold the arrow key, fly through 300 matches in 10s, no spinner in sight. (1x speed)
Does anyone want to make $50 by letting us watch them use Macro for the first time?
Requirements:
-Has never used Macro before
-Is willing to connect a gmail/workspace account
-Can share their screen
-In the US
-Has Venmo
Here's the link to book!
https://t.co/O3VhO4YN9p
Cool milestone for us, we just passed 100 stars on Github! Here's the repo if you want to follow along with our development: https://t.co/mlGb1AEiNn
And a recent changelog...
- 📞 Calls: new frontend, recording, production transcription, audio/video selection, and noise filtering
- 🤖 AI improvements: MCP support + docs, email tools, better tool rendering, and UI for failed tool calls
- 🔎 Smarter search: channel message entity, tab filters, profile pics in filters, and improved Command-K
- 💬 Channels: scroll stability, thread reply improvements, and email invites for new users
- 📧 Email: async Gmail ops worker with rate-limit retry, plus task creation from threads & messages
- 📱 Mobile/iOS polish: action drawers, drawer sizing, portrait lock, and better touch handling
- 🎨 Fresh look: updated Macro Dark/Light themes
We're Calling It a Soft Launch
After years of building, Macro is officially open to the world today.
We're calling it a "soft launch", because we're being honest. The product isn't perfect yet. But it's already better than the patchwork of tools most engineering teams are stuck with, and we think it's time for other people to try it.
The Problem We Couldn't Unsee
If you're an engineer at a startup, your workday probably looks something like this: Slack in one tab, email in another, Notion in a third, Linear in a fourth, Google Docs in a fifth, and ten more miscellaneous tabs lined up after that. You spend hours just context-switching between tools that were never designed to work together.
We tried everything. Slack was fine for chatting but terrible for anything structured. Notion was powerful but sluggish and never felt like it was built for engineers. Linear was clean but only solved one piece of the puzzle. Superhuman made email faster but didn't solve the fundamental problem: your email still lived in a completely different universe from your tasks, your docs, and your team conversations.
None of them are bad products. They just never added up to something cohesive. You'd stitch together six or seven tools with integrations and workarounds, and the result was always a fragmented mess that ate your focus and slowed you down.
We got tired of it. So we built Macro.
One Tab to Close Ten
Macro is the unified workspace — email, docs, tasks, team chat, canvas, code, and agents, all in a single interface. Same text editor everywhere. Same keyboard shortcuts. Same permissions model. One search that reaches across everything.
The backend is Rust because we wanted something fast and we wanted a codebase that could scale without becoming a maintenance nightmare. The source code is available, so you don't have to worry about being locked into a tool you can't inspect or extend.
We use Macro to build Macro. Every channel message, every doc, every task, every email at our company runs through the product. Our greatest strength is that we're shipping the tool that we reach for every single day.
Why "Soft Launch"
Building an entire office suite from the ground up is an enormous undertaking. There are still rough edges and some features are more polished than others. We're going to ship improvements fast, and some things will break along the way.
But even in this state, Macro is a fundamentally better way to work than juggling Slack, Notion, Linear, Gmail, and whatever else you've duct-taped together. The unified experience alone changes how your team operates. When your AI assistant can search across your emails, channels, and documents in one shot, when you can reference a task from a channel message, or draft an email while staring at the relevant doc , that's a different and better way of working.
We're not waiting until everything is perfect to let people in. We'd rather build in the open, get real feedback from real teams, and iterate from there.
Try It
If you're an engineer or run an engineering team and you're tired of the tab graveyard, come try Macro. While it's not perfect, it's available today.
We built this for ourselves and now we're building it for you.
→ https://t.co/SjwkwFOsuw
right now everything in the world is telling you to go faster, ship more, add that feature, start another project
so i'm actively working on feeling ok not doing any of that
Do it all- In Macro
Macro is the unified system for engineering teams. It brings together your messages, agents, tasks, docs, PRs, and emails in one interface.
Engineers are currently drowning in tab chaos. Slack in one window, Notion in another, Gmail in a third. Every context switch costs focus and momentum. Macro is being built to eliminate that friction by collapsing the tools engineers actually live in into one fast interface.
Join us in Flatiron this Wednesday 3/25 — we have a new screenprinter in the office, so bring some blanks to print cool designs on. And co-work, and see what we've been up to. https://t.co/TAHNUyRVyj
Loro Mirror is released! It makes building a collaborative experience much easier. You declare a simple schema and keep using setState. Mirror syncs a typed, immutable view of your app state with CRDTs, so you avoid most of the diff/patch glue