Atman

@AKBanga @JACCJournals @SaurabhSDani @usman_aliakbar It gives me a not found error.

🚨 Excited to share our latest publication in @JACCJournals - JACC: Advances! 🚨 Our study compares #Inclisiran vs #Ezetimibe in CAD + hyperlipidemia showing lower 1-year mortality, MI, and stroke with inclisiran, along with better LDL goal attainment: doi.org/10.1016/j.jaca…

They are now refusing to state the number of American casualties in the Iran war. When regimes do this, it is a HUGE RED FLAG.

claude mere bhai saas toh lele

@mutant1879 @mutant1643 What explains the difference? Harness? Model smartness? Tool call?

Just ran Codex and Claude Code on a FreeGym internal benchmark. Claude Code scored 66/100, while Codex came in at 38/100.

Even Goldman is tracking the mixed messages.

Google has several legends two of them: Jeff Dean: just incredible and always make sures he does right by people and world Demis Hassabis: just incredible at the vision he realized in the world and the foresight. I feel like they are the best humans among us on long horizon self play. In the sense they literally established the pillars of the current event horizon we live in. (I promise this is the last glazing post but wanted to make my statement)

"gremlins" and "goblins" are the new "you're absolutely right"

One emerging phenomenon with claude code I see - it is writing, verifying, executing complex code faster than we can understand. Since most stakeholders care about "getting the results" or "smth that works", engg doesn't know what the code is really doing anymore; feature or bug?

@Shreyagupta08 I think about this a lot, and it scares me sometimes. Everyone wants to ship, but who’s going to own the outcomes of people writing code they don’t understand? Something about it feels wrong. Because beyond a certain point, if you don’t know what you’re doing, you’re in soup!

@markgurman fanboy here but I’m sure that 26 minutes must not have been an easy squeeze? macs already have amazing battery life to begin with

Only with Apple is the new MacBook Pro lasting an extra 26 minutes - after a year and a half - considered an “astonishing” upgrade.

From this morning's Financial Times comments section. It's hard to disagree...

This is a brilliantly silly way of illustrating something real: the inflation of labour intensive, locally delivered products and services vs quality adjusted deflation in technologically advanced globally mass produced goods… a spot of the Baumol effect here, familiar to anyone who has studied economics too ….

@chatgpt21 Would not use it purely basis privacy concerns.

The $20/mo AI standard is fracturing. I’ve been watching the pricing floor for LLMs, and I had gotten internal word a few days ago that Alibaba Cloud was going to release a subscription that was going to force a lot of hands. They’ve just officially launched an AI Coding Plan that effectively cuts the cost of Pro-grade coding assistance by roughly 85%. I actually thought it was going to be $5/mo, but they just released it at $3/mo* for the first month. 15$ if you want pro 😭 For that $3/mo entry point (Lite plan), they are providing a massive quota (18,000 requests/mo) across a suite of models. We’re seeing: • Qwen-2.5-Coder & Qwen-3.5-Plus: Currently some of the strongest models for Python and C++ with massive context windows (up to 1M tokens). • Full compatibility with Cline and Claude Code • Access to the new GLM-5, Kimi K2.5, and MiniMax M2.5 all under one API key.

@slow_developer I think it is a function of model capability and training costs. Output price rising faster than input may be indicative of more internal thinking.

google keeps raising gemini flash-lite prices with every release: 2.0 flash-lite: $0.075/1m input, $0.30/1m output 2.5 flash-lite: $0.10/1m input, $0.40/1m output 3.1 flash-lite: $0.25/1m input, $1.50/1m output but the 3.1 flash-lite price makes more sense since it's basically a distilled version of gemini 3 pro

@JeffDean Google has done a stellar job of making high-quality intelligence available at a fraction of the price. Excited for the applications this enables!

⚡ Excited to announce Gemini 3.1 Flash-Lite! We’ve set a new standard for efficiency and capability to give developers our fastest, most cost-effective Gemini 3 model yet. We engineered this model with thinking levels, allowing it to handle high-volume queries instantly, while scaling up its reasoning for complex edge cases. By the numbers: ⏱️ 2.5X faster time-to-first-token than 2.5 Flash while being significantly higher quality 📉 $0.25 per 1M input tokens 📊 1432 Elo on LMArena & 86.9% on GPQA Diamond Thrilled to see what developers build with this kind of speed and quality at scale. Available now in Google AI Studio and Vertex AI. blog.google/innovation-and…

This lady ate a surface-to-air missile, sustained a 12-14 G rocket-propelled ejection, likely hit the ground at 15+ miles per hour… And then she popped up smiling at the locals who came to help her. Tell me again that women don’t belong in combat.

@praeclarum @KooKiz what a legend

@KooKiz I help balance it out by using requests like "good morning"

It has been said already, but Copilot pricing is completely broken. Use it while it lasts.

@runItsthekid have done a marvellous job at Abracadabra! So stoked to have y'all put out new music.

@SiddhiiShahh Happens…. darling

@HeyAtman Bro what! I suppose you knew this, you are a tech genius

I was today years old when I realised Apple Watch charger could charge Airpods.