Sponsored by
Welcome back to The AI Field.
The big labs are done just shipping models. OpenAI just unveiled its first custom chip. Anthropic put Claude inside your Slack as a real teammate. And ByteDance's new Seedance 2.5 generates a clean 30-second AI video in one shot, no stitching.
In this issue:
🌶️ OpenAI built its own chip with Broadcom
💬 Claude Tag turns Slack into an AI teammate
🎬 ByteDance's Seedance 2.5 breaks the 30-second video barrier
Plus: OpenAI's cyber play
Read time: 5 minutes.
Scale your customer experience across all channels
Most teams scale customer service by hiring, but with ElevenAgents, real-time voice and chat agents handle high volumes of interactions and resolve them faster.
When routine requests come in, they get answered in the moment, in 70+ languages. However, the complex ones route to humans with full context, so your team focuses only on the conversations that need judgment. The experience holds up because the agents sound human, not robotic. And with Expressive Mode, they read tone and adapt - de-escalating, reassuring, and guiding each conversation to a clear resolution.
You can train every agent on your knowledge base, SOPs, and policies, deploy across voice, chat, and WhatsApp, then keep improving with real-time CSAT tracking, A/B testing, and Guardrails.
The outcome is the one leaders care about: higher resolution rates, higher CSAT, and support that scales with demand instead of headcount. Pricing is transparent and flat at $0.08 per minute, with no add-ons to work around.
The AI Field: OpenAI just unveiled its first custom chip, an inference processor called Jalapeño built with Broadcom. It's a clear move to own more of its stack and lean less on Nvidia for serving its models.
Key details:
Jalapeño is a reticle-sized ASIC designed from scratch around LLM inference, and it hit tape-out in about nine months, a remarkably fast cycle.
OpenAI used its own models to help design the chip. Lab samples are already running workloads, including GPT-5.3-Codex-Spark, at target power.
It's the first of a multi-generation platform with Broadcom and Celestica, set to deploy at gigawatt scale starting late 2026.
Why This Matters: Whoever controls the inference hardware controls the cost and margins of AI. By designing its own silicon, OpenAI is trying to escape Nvidia's pricing and build the "full stack." The model makers are becoming chip makers.
Anthropic
💬 Claude Tag Turns Slack Into an AI Teammate
The AI Field: Anthropic launched Claude Tag, which drops Claude into your Slack as a persistent team member. You @-mention it, hand off a task, and it works through the steps using the tools you've connected, then reports back in a thread.
Key details:
It's "multiplayer," one Claude per channel that everyone can see and pick up from, and it learns your team's context over time so you stop re-explaining things.
With "ambient" mode on, it proactively flags things and chases stalled threads without being asked.
It's in beta for Enterprise and Team customers. Anthropic says 65% of its own product team's code is now written by its internal version.
Why This Matters: This is the shift from "AI you prompt" to "AI that sits in your workflow and acts." Putting an agent where work already happens, your chat, is how AI actually gets adopted. The AI coworker just clocked in.
The AI Field: ByteDance's new Seedance 2.5 generates a single 30-second AI video clip in one shot, no stitching, complete with scene changes and synced audio. Rivals like Sora 2 and Veo 3.1 still top out around 8 to 15 seconds.
Key details:
It co-generates video and audio in one model, holding character looks, lighting, and motion steady across the full clip, in 4K.
It can take up to 50 reference inputs (images, audio) at once, useful for multi-character scenes.
Announced in global enterprise beta, with a public launch targeted for early July 2026.
Why This Matters: Length and consistency were the two things keeping AI video out of real production. Cross those, and "AI video" stops being clips and starts being scenes. The gap between a demo and a usable shot just narrowed a lot.
Four tools on our radar that are worth checking out.
🤝 Lindy - build AI agents that handle your email, scheduling, and repetitive workflows, free tier.
🖼️ Recraft - an AI design tool for logos, icons, and on-brand visuals, free tier.
⚙️ Gumloop - drag-and-drop AI automations for the busywork, no code, free tier.
🎙️ Descript - edit video and podcasts by editing the text transcript, free plan.
🔥 THE AI FIELD'S TAKE
Put these stories together and the pattern is obvious: the AI labs are done being just model companies. They're building the whole stack.
OpenAI is making its own chips so it controls the cost of running AI. Anthropic is pushing Claude off the chat window and into your Slack, where the actual work happens. And the video models are crossing from clips into real scenes, coming for Hollywood's storyboard.
The first phase of AI was a better answer box. This phase is different. The winners are building the silicon underneath, the agents in the middle, and the studio on top, all at once. The model was never the whole game. It was the foothold.
OpenAI launched Daybreak, its cybersecurity push. GPT-5.5-Cyber, a Codex Security plugin, and "Patch the Planet" aim to auto-find and fix software flaws, with government partners.
Yann LeCun called xAI "kind of a failure." All 11 non-Musk co-founders have left, and he warned the whole industry is heading for a bubble.
📲 TRENDING IN AI
OpenAI's first chip lit up X, mostly framed as the opening shot in an "OpenAI vs Nvidia" fight over who powers AI.
Seedance 2.5's 30-second clips went viral fast, with creators posting side-by-sides against Sora 2 and Veo 3.1.
Claude Tag got developers talking after Anthropic said its own agent now writes 65% of its product team's code.
💡 PROMPT OF THE WEEK
Turn a goal you keep avoiding into motion.
Tool: ChatGPT, Claude, or Gemini
Prompt: "Here's a goal I keep procrastinating on: [your goal]. Break it into the three smallest possible first steps, each doable in 20 minutes, then ask me which one I'll start with right now."
What you'll get: a goal that suddenly feels doable, and a clear first move instead of a vague to-do.
👀 WHAT TO WATCH
Gemini 3.5 Pro: still in limited preview, with prediction markets at roughly 50 to 55% odds it ships by June 30.
Seedance 2.5: the public launch is targeted for early July, when everyone gets to test the 30-second claim.
Fable 5's return: prediction markets put it around 57% to be back before July 1.
WRAP-UP
That's all for today's roundup!
🌶️ OpenAI unveiled Jalapeño, its first custom AI chip, built with Broadcom.
💬 Anthropic's Claude Tag puts an AI teammate inside your Slack.
🎬 ByteDance's Seedance 2.5 generates a clean 30-second AI video in one shot.
What did you think of today’s edition?
Until next time!
Olle | Founder of The AI Field