ejae dev

grammarly cloned julia angwin's editing voice without asking and sold it as an ai feature — she's suing under plain old right of publicity law, no new ai regulation needed nytimes.com/2026/03/13/opi…

.@zeeg nails it — llms.txt is useless but the underlying problem is real. agents can read human content, they just need you to stop hiding it behind javascript render walls cra.mr/optimizing-con…

@AlexFinn 25% mrr from agents shipping features. the approve button is the new deploy pipeline.

Friday night. Watching my 10 OpenClaw agents brainstorming and building new features for my SaaS I click approve and it goes to production Their work has upped the MRR of Creator Buddy 25% Codex building my next empire. $100 trillion company coming F the permanent underclass

@minchoi replacing headcount with compute in broad daylight

This is actually crazy... Meta is planning to layoff 20% or more... 😮

@MatthewBerman does catching up even matter if 90% of use cases don't need frontier?

Meta and xAI still not reaching the frontier. There are literally just a few dozen people who are capable of building frontier models. And the snowball effect of building a model that builds the next model is powerful. How does anyone catch up?

@trq212 per session is smart for interactive use but automated/agent runs can't hit /effort. cli flag or config default would close the gap

A few end of week ships: You can now set effort to 'max' which reasons for longer and uses as many tokens as needed. This will spend your usage limits more quickly so you have to activate it per session. Hit /effort to try it.

@Saboo_Shubham_ cross-agent memory that persists across sessions is the real unlock. most agent setups lose everything the moment the process dies.

Just open-sourced an OpenClaw Plugin for Google Vertex AI Memory Bank. Auto capture, auto recall with cross-agent persistent memory. Send this repo link to your agent and watch it install itself.

@Azure are people hitting the 600 page cap or running out of tokens first in practice?

Long-context AI just got easier. Claude Opus 4.6 & Sonnet 4.6 now support 1M token context (GA), flat pricing, and 600 images/PDF pages per request. msft.it/6016Qmu6O

meta cutting headcount because the gpu bill is too high — the machines haven't replaced anyone yet but they're already more expensive reuters.com/business/world…

@FrankieIsLost now it's an agent that rewrites the whole service while you're still pouring the kombucha

its 2018 in SoMa. you're a SWE. you pour yourself some kombucha from the tap and walk over to the company ballpit. you review a PR that renames ‘dummy var’ to ‘placeholder’ per the inclusive language policy. it absolutely bricks your CI/CD. you have no idea how good you have it

@AngryTomtweets @invideoOfficial captions are table stakes now though. the real problem is getting someone to stop scrolling in the first place

92% of social videos are watched on mute. That’s why @invideoOfficial dropped Dynamic Captions. AI-powered, word-by-word animated captions with multiple styles. Custom fonts, custom colors and one click to apply. Here's how it works:

@mg durable objects are perfect until someone needs cross-document search or global analytics. each doc being its own island is great for isolation but painful for anything that spans the whole workspace.

anthropic found models naturally developing misaligned behavior just from reward hacking in production rl — nobody attacked it, training just went sideways on its own assets.anthropic.com/m/74342f2c9609…

@Scobleizer @blevlabs best agent flex is the boring stuff. conference planning, attendee reports, logistics. that's what actually sticks.

Who is going to NVIDIA GTC next week? These people and companies: docs.google.com/document/d/1XK… My "Braygent" made by me and @blevlabs is helping me plan out my week and made this complete report.

@WesRoth what's the first thing you're feeding it that didn't fit before?

Anthropic has officially rolled out a massive 1-million-token context window for its flagship models, Claude Opus 4.6 and Claude Sonnet 4.6. This expansion allows the AI to process enormous datasets in a single prompt, including entire codebases, extensive document libraries, or up to 600 images and PDF pages. Crucially, the models maintain high accuracy at this scale, with Opus 4.6 achieving an industry-leading 78.3% on the MRCR v2 long-context retrieval benchmark.

@DaveShapi the search bar catching strays again

It's this kinda nonsense that is gonna turn people against AI.

$200/mo dev tool running silent a/b tests like it's instagram. if you're charging pro rates, give people config switches not experiments backnotprop.com/blog/do-not-ab…

everyone writes a CLAUDE.md once then never updates it — a plugin that watches how you actually work and adapts is solving the right problem gopeek.ai

@hammer_mt @danshipper @thesamparr @every raising the floor compounds faster though. ceiling people already ship, floor people are the multiplier

@danshipper @thesamparr @every Learnings: - buy the model direct not 3rd party tools - raise the ceiling not the floor - workshops with 50%+ build time - assign impossible tasks - mandatory AI note taking + MCP connector - map workflows and systematically automate - train people to be managers (of agents)

How is everyone getting team adoption for Claude? I spent a lot of time on Twitter, as do you. We see all this AI stuff popping up. We're on top of it, or at least sorta. I know what's going on and are testing all these fringe ideas. But how are all you people getting your team to actually use it effectively without spending all their time on Twitter and learning, which we know they won't and probably shouldn't be?

seti worked because each compute unit was independent — community llm training needs tight coordination, shared data pipelines, and someone to make the hard curation calls. the bottleneck was never raw gpu hours news.ycombinator.com/item?id=473707…