Hermes Agent NEW Super-App and DeepSeek v4 Catches Up To Opus 4.8?
TL;DR
DeepSeek V4 is getting real traction in production: Flo Crivello of Lindy says they moved 100 percent of traffic from Anthropic to DeepSeek V4, saving millions while seeing better performance on some core agent tasks.
The pricing gap is brutal for closed models: Riley compares DeepSeek V4 Pro at about $1.30 per 1 million input plus output tokens with Anthropic Opus 4.8 at $30 and OpenAI GPT 5.5 at $35, framing DeepSeek as 23x to 27x cheaper.
Codex and Cursor are both becoming internal app builders: OpenAI's new Codex Sites and Cursor's Canvas let teams generate and deploy internal tools connected to Gmail, Slack, and company data, which Riley sees as the first step toward full AI coding super apps.
Cursor may be aiming at a GitHub-for-AI future: Riley reads Canvas plus new public profiles as clues that Cursor wants a world where people share, fork, and remix AI-built apps, not just code repositories.
Apple may be turning iMessage into an agent store: Poke is now an officially approved iMessage agent with gray Apple-style messaging, and Riley thinks WWDC could bring an approval path for companies to ship agents directly inside Messages.
Hermes is emerging as the open super-app alternative: Hermes Desktop replaces the old terminal-heavy setup with a polished app that supports multiple model providers, built-in skills, memory, messaging connections, and artifact management.
The Breakdown
DeepSeek V4 may already be good enough to replace Anthropic and OpenAI for some agent workloads at 23x to 27x lower cost, and Riley argues that cost gap is about to reshape the whole AI agent market. He also tracks a wild week of product moves, from Codex and Cursor turning into internal app builders to Apple quietly turning iMessage into what could become an agent platform.
Was This Useful?
Share
Keep Reading
Make Alcreon Yours
Tune your feedFive quick questions, and the feed ranks what matters to you first.Or just get notified
The weekly Echo. Signal worth keeping in your inbox.
Every new piece, announced on X.
Read Next
See all
Playbook
Cheap Models, Hard Tasks
Most agent workflows route every step to the frontier model by default. The bill scales with how chatty the agent gets, even when most steps don't need that brain.

Playbook
Tasteful Skills
“Tasteful Skills” argues that the best agent skills are not documentation or best-practice lists.

Playbook
The Art of Tasteful Prompting
Learn how tasteful prompting helps you move beyond generic AI output by shaping context, style, and judgment from the start.