Railing in and actually optimizing AI token spend for engineering teams is going to soon become a massive industry priority If you want to understand context window burn, token waste due to compaction and more @EntelligenceAI has it covered

Got minee finally!

we're hiring a marketing manager in SF and the job is basically: run weird experiments, break some, double down on the ones that work. not a content-calendar role. $120–150k Esops. in-person. Apply through the link 👇

Been saying this since day 1. Check out research.entelligence.ai - for a very detailed analysis on the code <> debugging ratio token spend. For every $1 spent on generating code, $2.4 is spent on fixes and debugging

just got my AI wrapped today 🤩 1M tokens burned and my wrapped called me a Falcon 🦅 peak activity time 2 am 💤 niceness score that doesn't exist 👀 it knows me too well 💀 get yours wrapped.entelligence.ai 👈🏻

These really look cool! What do you guys think?

We have dashboards for everything now, yet most of us can’t answer basic questions about how we’re actually using AI. Kinda strange when it’s becoming part of our workflow every single day. Interesting idea from @EntelligenceAI wrapped.entelligence.ai/

17.2M tokens. 5 sessions. 6.5 tools per turn chaos. Entelligence called me a whale 🐋 Spotify Wrapped wishes it had this level of damage

"we burned 20B tokens this month." okay and? did it ship something great? or did it ship a very expensive bug? because the count can't tell you. it never could! it only measures how much the model talked, not whether it said anything worth hearing. - volume is noise - behavior is signal. @EntelligenceAI tracks what the model actually does, not how much it spends doing it. check now: wrapped.entelligence.ai/

We have Datadog for infra. Linear for issues. Figma analytics for design. But for the tool engineers now spend the most time in? Nothing. AI coding visibility is the gap nobody's talking about. @EntelligenceAI just launched the first step to fix it and it's actually fun. 212M tokens in. Rarer than 81% of agents. Beaver confirmed. wrapped.entelligence.ai

"Using it well" is actually a skill to optimize on Most engineers don't yet understand the differences between context pollution and context pressure, how to optimize their usage across different tools and what updates can be made to agent.md files to optimize it

So so much cooking @EntelligenceAI Determined to make sure coding agents and reviewers have all the production context they need to never make the same mistakes more than once

what's your spirit vibe coding animal... DM to get one :P

The @EntelligenceAI team killed it yesterday We invited several of our customers to come race Porsche on the pier Bonding out of the office is nearly always one of the best ways to build the relationships that actually matter

Cookinggg ⁦@harsh_logs⁩

Porsche racing today!! Yes that’s the @EntelligenceAI logo on a Porsche. Kinda crazy sometimes the kinda things you can do 😆 Get away from your laptops and come join for the day! Luma below

👀 somethin cooking...

The stat that jumped out to me wasn't code churn. It was that only 21% of review comments get addressed. That's like paying for a map and ignoring the directions. Excellent analysis of where engineering time actually goes.

This lowkey matches my observations Code slop is so bad even in my code at work, that I literally waste one more prompt on deslopping Unnecessary tokens wasted everyday, and this gets worse with every new model Nice to see someone solving this

Everyone is talking about how AI is making developers more productive. But productivity and progress aren't always the same thing. After analyzing 1M pull requests across 2,444 organizations, Entelligence found: • 1 in 4 lines of code gets thrown away • Only 21% of review comments are addressed • Reactive work dominates engineering time at many teams The takeaway: The challenge isn't generating more code. It's making sure more of that code actually ships and creates value. Interesting research from Entelligence.