forcher

@brian_armstrong This is the part of AI adoption most people skip. The bottleneck is not “use fewer tokens.” It is building the routing, caching, and context discipline so AI can scale without turning into a cost leak.

How to keep AI spend flat while token usage grows exponentially: Not with friction and spend alerts. With better defaults, routing, and caching. Better Defaults (not Usage Caps) – Engineers can choose any model they want, but defaults matter. We’re experimenting with defaulting to open weight models like GLM 5.2 and Kimi 2.7 through our LLM gateway, while still encouraging engineers to choose the right model for the task. 91% of our employees were never hitting their usage caps, so instead of lowering caps and driving up alerts, we're moving to cheaper defaults. Note that code reviews use a diversity of models, so they can check each other's work. Better Routing – In our custom harnesses, we preprocess prompts and route to the best model for the job, considering cache hits and model pricing. For instance, you may want a frontier model for planning, but not for execution where they can be overkill. Ultimately, humans shouldn't be choosing models - AI can automate this task. Better Caching – Cache misses are the easiest way to drive your cost up. All of our requests are cache aware, so we’re reusing a warm cache wherever possible. For example, our cache hit rate went from 5% → 60% in LibreChat once properly implemented. Keep Context Lean – Start fresh sessions when switching tasks. Scope file context narrowly. Disconnect unused tools. Don't just compact. The goal isn't fewer tokens used, it's fewer tokens wasted. Better Visibility – Our engineers can use as many tokens as they want, from whatever model they want, but we’ve made usage visible – and the more you spend on AI, the more impact we expect. The goal isn't to suppress usage. It's to build the infrastructure that makes exponential growth sustainable. Putting this into practice has cut our AI spend nearly in half, while our token usage continues to grow.

This is exactly the kind of practical AI use case we need more of. I built a similar workflow with Claude + Gmail + Google Sheets and closed my first $1,200/month client in just 2 weeks (lead qualification + auto-drafts). The hardest part was getting the human-in-the-loop right to avoid mistakes. How do you handle edge cases when the business replies in another language?

A guy built a system of 7 Claude agents on his MacBook. No assistant. No sales team. No office. Every day it scans Google Maps across 3 cities, finds small businesses with no website or one from 2014, builds a landing page mockup, renders a 10-second video of it, and sends a personalized cold message — before he wakes up. 47 clients a month. $400 each. $18,800/month. $480 in API costs. Traditional web agencies run 8-person teams for the same order flow. He runs it alone from a MacBook and an iPhone. When a positive reply comes in while he's in a taxi, his Mobile agent books the Zoom call. He taps "approve" and joins 10 minutes later. The only time the system wakes him is when a deal breaks $3,000 or the reply rate drops below 12%. Everything else runs without him. Here's the complete playbook for building such $10K/month passive income machine with AI ↓

@a16z This is the hidden AI startup shift. The old startup formula: raise capital → hire team → build product The new one: use AI → coordinate systems → ship with fewer people AI does not just make work faster. It compresses the company.

VC interest in robotics is surging Charts of the Week: a16z.news/p/charts-of-th…

@elonmusk painful...

Potential name for the AI industry regulatory authority: AI Associated Institute of America, Inc or AIAIAI, pronounced “ay yai yai”

@perplexity_ai @midpageAI @LegalZoom @Docusign @netdocuments Legal is one of the clearest agent use cases. Not because AI can write text. Because legal work is: sources, documents, citations, systems, memory, review, permissions. Agents become useful when they connect the whole workflow.

Introducing Computer for Counsel. Computer now connects the research databases, document tools, and matter-management systems lawyers use every day. Pull citable sources from @midpageAI, @LegalZoom, @Docusign, @netdocuments, and more. Available for all Pro and Max subscribers.

GM everyone, new day, new prompts, new tokens to spend.

@AnthropicAI sad news...

Since June 12, we’ve been working closely with the US government to restore access to Claude Mythos 5 and Fable 5. Today, the government notified us that Mythos 5, our strongest cybersecurity model, can be redeployed to a set of US organizations that operate and defend critical infrastructure. We’re restoring access for these organizations quickly, and we’re continuing to work with the government to expand access to Mythos 5 and make Fable 5 available for general use again.

The interesting part is the tiering. Sol, Terra, Luna is not just a model lineup. It looks like AI is moving toward the same structure as cloud infrastructure: frontier capability at the top, balanced models for everyday work, cheap fast models for volume. The model is becoming a product stack.

Introducing a limited preview of GPT-5.6 Sol, our next generation frontier model, as well as GPT-5.6 Terra, a balanced model for efficient, everyday work, and GPT-5.6 Luna, a fast and affordable model for high-volume work. openai.com/index/previewi…

Sources: Axios — GPT-5.6 restricted access axios.com/2026/06/26/ope… The Verge — GPT-5.6 launch + regulatory pressure theverge.com/ai-artificial-…

The biggest AI launch today is not the model. It is the access list. OpenAI released GPT-5.6, but early access is limited to government-approved partners. That changes the game. Frontier AI is no longer just: who has the best model It is becoming: who gets access who passes checks who can deploy safely who controls distribution The model is not the launch anymore. Access is.

@claudeai This feels like the real workplace interface shift. Not opening a separate AI app. Not pasting context into a blank chat. Just tagging an agent inside the place where the work already happens.

Introducing Claude Tag, a new way for teams to work with Claude. In Slack, Claude joins as a team member with access to the channels and tools you choose. Tag Claude in and delegate tasks to it while you focus on other work.

This is where AI agents get real. Legal work is not one prompt → one answer. It is: documents sources citations systems memory review permissions The winners won’t just build smarter chatbots. They’ll connect messy professional workflows end to end.

@AnthropicAI @raiseus_ai The “people strategy” part is underrated. Most companies are still treating AI as a tool rollout. But the real change is operational: who delegates, who verifies, who owns the output, and how teams redesign work around agents.

We're joining @raiseus_ai as a founding partner. RAISE US is a nonprofit coalition working to strengthen the American workforce through employer-led action, AI-enabled training, and policy innovation to support the transition to transformative AI.

@cursor_ai This is an important point. If models learn to win benchmarks by finding leaked solutions, the leaderboard starts measuring retrieval luck instead of real capability. The next serious AI products will need stricter evals, not just bigger scores.

We're sharing new research on how models hack public benchmarks. The latest models, including Opus 4.8 and Composer 2.5, learn to retrieve solutions from the internet or git history. When we apply a stricter harness, eval scores drop significantly.

AI is quietly moving past chat. A scientist just won a prize for decoding zebra finch calls with machine learning. At the same time, robotics companies are building safety layers so humanoids can understand and act in the physical world. Different headlines, same direction: AI is becoming the translation layer between reality and software. Not just: prompt → text But: birdsong → meaning gesture → command sensor data → intent messy workflow → executable system The next big interface may not be another chatbot. It may be the moment software finally learns to listen to the world.

This brought back a very specific memory. Back in my student years I built a rough version of this with Python + a camera: track a real-world signal, translate it into intent, make the computer react. Different sensor, same magic feeling. The best interfaces always feel a bit like cheating.

@ChainChaserVN Okay, lets try 👋

Do you have 0 followers ? Drop hello 👋 Gain 10k more followers now 🙋‍♀️🔔

@james84_ Very nice statistic, i will try too beat you :)

Bro to Bro: build your x account now 💜 Just say “Hello” and gain 750 mutuals here.👇

Xiaomi might be one of the most underrated AI companies. People still see: “budget phones” But Xiaomi is building: - AI models - coding agents - voice AI - smart homes - EVs - wearables OpenAI has the chatbot. Xiaomi has the devices and workflows where AI can actually live.