Unlock Revenue with Agentic AI and Claude Code

🚀 **Zed's new Agent Stats dashboard reveals AI agents are exploding in code editors – with Claude Sonnet's p90 latency spiking 44% in just three weeks!** We've been diving deep into this public weekly data from Zed, tracking anonymized session counts, turn volumes, and response times across models like Claude and GPT variants. . Key insights we're applying in our AI workflows at Digitly: • Monitor **p10/p50/p90 latencies** to pick agents that keep devs in flow, not waiting. • Leverage Zed's Agent Panel for code gen, refactoring, and debugging with tool calling and checkpoints. • Queue messages smartly – Zed batches them at turn boundaries for smoother external agent interactions. . In one project, we observed how deeper session engagement (like Claude Agent's edge over natives) slashed iteration time by streamlining multi-turn debugging. . Boost your productivity: Check Zed's metrics at zed.dev/agent-metrics and tweak your stack today. . Comment below: Which AI agent's latency bugs you most? #AIAgents #CodeAutomation #DevTools

I've been using Claude with artifacts and memory persistence more lately, and it's got me thinking about what makes AI workflows actually stick. The thing that surprised me isn't the memory part, Claude already remembers our conversations pretty well. It's how much more useful it becomes when you stop treating it like a search engine and start treating it like a thinking partner. I used to ask Claude: "How do I set up a webhook?" Now I ask: "I'm building this automation flow, here's what I'm trying to solve, what's the cleanest approach?" Then we iterate on it together over multiple sessions. The difference is huge. Instead of getting generic tutorials, I get solutions that build on our previous conversations. Claude remembers the constraints I mentioned last week, the tools I prefer, even the specific quirks of my setup. What really clicked for me was realizing these AI tools work best when you bring them into your actual problem-solving process, not just the "how do I code this" moment at the end. I'm curious, how are you all structuring your interactions with AI tools? Are you finding certain conversation patterns work better than others? #AIWorkflows #Automation #AITools #BuildInPublic #Claude

The bottleneck has moved. While coding speed used to be the hurdle, AI assistants have flipped the script, flooding teams with shippable code while release tools still lag behind. I'm thrilled about our 5 new releases because they tackle the real friction points faced by our customers daily: ➡️ Coordinating multi-service dependencies ➡️ Segmenting feature releases based on sensitive data ➡️ Managing data warehouse changes alongside apps. Harness is bridging the AI Delivery Gap. Dive into the full details here: https://lnkd.in/eXSvQ4tC What do you think is the biggest challenge in keeping up with AI-driven release speeds? Let's discuss! 🚀

Stop listening to the AI hype and start looking at the architecture 🔎 While the rest of the industry was reacting to the AI boom and building surface level chat-bots, we were in the kitchen cooking up something unprecedented. In this 4-part series, our Chief Product Officer Matt MacInnis and Head of Solutions Consulting Brian Mazzaferri sit down to reveal why and how we built Rippling AI–and how it’s shaping the future of work. From unlocking cross-product answers to building a permissions model that actually respects your data, Rippling AI is redefining business insights. The full story drops over the next week. Got burning questions? Our Chief Product Officer Matt MacInnis is answering them all LIVE on Reddit on 4/14. 👇 https://lnkd.in/eKmz4Sfg

Moving forward, we're building for AI agents as much as for people. But especially for people delegating work to AI agents. In the coming days, we're releasing an integration to Claude Cowork - the current leader in genuinely useful and capable agentic AI tools that can perform real work. And yes, it's really that simple... upload your project documents and tell it to "set up my project". It will ask questions about which of your Bluerithm templates to use, make suggestions, and get approval for things along the way, but our new AI-friendly backend services, + Claude-specific integration + Cowork = absolute game changer. More to come on this including end to end demos. 🤖

We're releasing an integration to Claude Cowork in the coming days. 1️⃣ Upload project documents 2️⃣ Tell it to set up your project 3️⃣ It will ask questions about which Bluerithm templates to use, make suggestions, and get approval for things along the way 🤖 #AIAgents #AI #ClaudeCowork #Commissioning

What if your client onboarding video updated itself automatically every time your product changed? That's the idea behind Remotion. Videos written in React code, rendered server-side. Fully parametrised. One component, infinite variations. Paired with AI, it opens up things like auto-generated product explainers, data visualisation videos, and personalised demo clips at scale. We've been exploring this as part of our AI automation work. The possibilities for businesses that regularly produce video content are significant. https://www.remotion.dev https://lnkd.in/gxCjPJP7

I came across a discussion where engineers talked about using AI agents in data pipelines. The pattern was clear. Everyone likes AI for small things, explaining SQL, quick scripts, debugging hints. It saves time and removes friction. But when it comes to larger systems, the confidence drops fast. No one wants an agent making changes in production on its own. Not because it’s useless, but because pipelines need consistency, and small mistakes don’t always fail loudly. So the real use case today feels simple. Use AI to move faster. But keep humans responsible for anything that actually matters.

Just finished using Weights & Biases; it helped reduce the development effort to setup AI Infra by 30% since it provides Infra + eval + RL in one stack, giving my team expanded set of tools to build, train, and deploy production-grade AI agents W&B Weave automatically tracks every LLM call using the @weave.op decorator, capturing inputs, outputs, costs, latency, and evaluation metrics without manual setup #CoreWeave

Voice AI solved latency. Now the real game begins. The new moat isn't speed — it's proving, statistically, which change actually moved the needle. → Phoneley handles millions of voice AI calls/month — 80% of callers can't tell they're talking to a machine → One script question changed → 5% lift in outcomes. Phoneley surfaced which one. → Edge: continuous statistical optimization of call results, not just answering phones → Bodis went SMB-first deliberately — brutal weekly feedback built the iteration engine before chasing enterprise → Bessemer found him via a LinkedIn post about 300-mile bike races, not a pitch deck The model is becoming a commodity. The durable position is the optimization layer above it. Full piece: https://lnkd.in/gKWcrqGb