Griffin AI September Update: GPT-5, Market Watcher & Community Wins

Griffin AI Team
 | 
Wednesday, September 3, 2025
Griffin AI September Update

Last month at Griffin AI new tools went live in the Agent Builder, the Agent Playground continued evolving, and the community brought in another round of inventive builds, showcased in the Community Gallery. 

Here’s a closer look at the highlights.

Network & Agents

This is where most of the movement happened: upgrades, a new AI Agent release, and an agent built for breadth.

Agent Playground and Community Agents glow-up upgrade

Agent Playground Glow-up Upgrade

The Community Gallery refresh!

The Agent Playground received a visual refresh this month. With the latest Agent Playground upgrade, avatars now sit neatly centered in the Playground and the light mode option makes longer sessions easier on the eyes. It shows that the space where people spend hours building is being shaped with care.

GPT-5 released in Agent Builder

GPT-5: Now live in Agent Builder

GPT-5: Now live in Agent Builder

The bigger change was under the hood. GPT-5 is now available in the Agent Builder, joining GPT-4.1. Builders can still rely on 4.1 when they need quick conversational agents. But when the job requires heavy multi-step reasoning or a workflow that needs to hold together across tools, GPT-5 is the better fit. It feels like the Builder is growing into a place where different models sit side by side, each pulled in for the job it suits best.

Launch of our Market Watcher Agent

Welcome to the Market Watcher Agent

Welcome to the Market Watcher Agent

The Market Watcher Agent arrived this month, a brand new agent for quickfire, accurate market reports. This new agent can scan two thousand tokens, surface short- and mid-term trends, and line them up against curated news. 

Used alone the Market Watcher works as a wide lens, spotting movement across the market. Paired with the Price Performance Analyst, which drills down into a handful of tokens in detail, it becomes a two-stage workflow: scan first, then study.

For analysts juggling dozens of assets, that’s a change you can feel in the way research gets done. For the full breakdown, check the blog post linked above!

Thought Leadership & Events

The conversation about where agents are heading also moved forward this past month.

Oliver Feldmeier on the future of AI agents

This month, our CEO Oliver Feldmeier laid out a powerful vision for the evolution of AI agents in an insightful X thread. He mapped out a three-stage journey from the "chat copilots" most people are familiar with today, to "governed agent orchestration" operating with strict security policies and budgets, and finally, to “truly autonomous, multi-agent teams”. This roadmap is a clear breakdown of the steps ahead for the entire industry. 

He explained that Griffin AI's heavy lifting is currently focused on mastering the second stage: building agents with tight, verifiable guardrails. However, this is not the final destination. Instead, every step and every learning from this work is a crucial building block for the third stage of truly autonomous agents. It's a forward-looking strategy that ensures the security and control we build today will enable the full autonomy of tomorrow.

Twitter Spaces with HTX

Oliver in the HTX Twitter Spaces

Oliver in the HTX Twitter Spaces

Oliver joined HTX Global for a great X Spaces session hosted by their X account this month. This session was a great spotlight for talking through what Griffin AI is building, answering questions, and setting out how our agents can now be part of crypto workflows with real utility. 

Oliver also unpacked some of the values at the heart of Griffin AI’s mission in this Spaces, such as protecting the rights of users to control their own data when accessing and using AI tech without being required to hand it over to the Big Tech companies who so often control the leading models currently. 

This decentralization of data ownership and control is central to Griffin AI’s mission and very much resonant with the founding values of blockchain. Check out the full replay above.

Community

The community side of things was as busy as ever at Griffin AI and our Build Smarter with Griffin AI campaign ran its course. With numerous agents entered in the competition, this was a highly popular campaign and saw great results, the best of which made their way into the Community Gallery.

Agent guidance and prompt pack released

Our prompt pack was sent out

Our prompt pack was sent out

These past weeks we have launched a series of posts and videos engaging with the community, highlighting great community-built agents, and showing how to use the Agent Builder and create better agents using our crypto-native toolstack. We also shared our set of detailed prompt templates with everyone who asked from the community on X. They weren’t flashy, but they worked: drop them straight into the Builder, tweak them a little, and you have a working agent. This was a great way of lowering the entry barrier for people just starting out, and it quickly fed into more experiments in the Builder.

Build Smarter with Griffin AI competition launch and winner announcement

The Build Smarter campaign

The Build Smarter campaign

The competition that followed saw great engagement from our users enthusiastically participating to compete. The challenge was clear: if hedge funds can trade with models and bots, why shouldn’t everyday users build their own? Submissions came in fast: the deadline was short, the prizes were XP, and the best agents got a spotlight in the Community Gallery. This turned into a real burst of building that pushed the Builder harder than ever.

The winners of the Build Smarter campaign were standout entries which used the tools available in the Agent Builder interface in a range of fascinating ways. Categories for entries in the competition covered DeFi, research, and multi-purpose agents. 

Airdrop Alpha Sniper won in the DeFi category and was designed for the degen crowd, scanning podcasts, forums, and news streams to pick out genuine airdrop opportunities before they’re widely known. 

Griffin AlphaTrader AI came first in the research category and worked in a different direction, taking on the role of a trading analyst that parses live data, runs pattern checks, and generates strategies in real time, all through a no-code interface. 

FunctionMapper AI triumphed in the multi-purpose agent category and stood apart as a utility for builders themselves, helping developers, designers, and product owners see what’s missing in their products, what features matter most, and where the next iterations should head.

Griffin AI is continuing steady in its course, of bedding down as a platform that grows through steady upgrades and a community that’s learning together, building and sharing faster each month.