# Anthropic's Latest- The Fight Towards AGI and How They're Measuring It

**URL:** <https://forum.zurb.com/t/anthropics-latest-the-fight-towards-agi-and-how-theyre-measuring-it/401>\
**Category:** Measuring design\
**Tags:** ai\
**Created:** [November 25, 2025, 5:01pm UTC](https://forum.zurb.com/t/anthropics-latest-the-fight-towards-agi-and-how-theyre-measuring-it/401 "2025-11-25T17:01:44Z")\
**Posts on this page:** 4\
**Page:** 1

<div class="post-metadata">

**Author:** ![ben](https://yyz2.discourse-cdn.com/flex034/user_avatar/forum.zurb.com/ben/32/442_2.png) [@ben](https://forum.zurb.com/u/ben)\
**Post date:** [November 25, 2025, 5:01pm UTC](https://forum.zurb.com/t/anthropics-latest-the-fight-towards-agi-and-how-theyre-measuring-it/401/1 "2025-11-25T17:01:44Z")

</div>

Anthropic is leading the way when it comes to programming. As a developer myself, I’m on top of these models.

Recommend reading their official post:

> **[Introducing Claude Opus 4.5](https://www.anthropic.com/news/claude-opus-4-5)**
>
> Anthropic is an AI safety and research company that's working to build reliable, interpretable, and steerable AI systems.

One thing that’s super interesting to me is how close all of these other models are in comparison. Wouldn’t that be extremely stressful? Knowing that the other competitors are a few percentage points away from not only dominating your expertise, but their whole ecosystem?

 ![image](https://canada1.discourse-cdn.com/flex034/uploads/helio/original/1X/83bda43308844f5b1bab24a0593dfd380b4983ec.jpeg)

I’m curious about when these incremental changes turn into an innovative, large leap forward, and what work that might consist of.

The other interesting thing, is how they’re _measuring_.

They reference this in their benchmarks [GitHub - sierra-research/tau2-bench: τ²-Bench: Evaluating Conversational Agents in a Dual-Control Environment](https://github.com/sierra-research/tau2-bench)

 ![image](https://canada1.discourse-cdn.com/flex034/uploads/helio/original/1X/a0bad679099ca098893b4061d5584672264e46c9.jpeg)

Are any of these means of measurement tools that other orgs should be using to identify gaps in their own tools? Or is this for only the core models?

---

<div class="post-metadata">

**Author:** ![nathaliesmith](https://yyz2.discourse-cdn.com/flex034/user_avatar/forum.zurb.com/nathaliesmith/32/443_2.png) [@nathaliesmith](https://forum.zurb.com/u/nathaliesmith)\
**Post date:** [November 26, 2025, 11:31pm UTC](https://forum.zurb.com/t/anthropics-latest-the-fight-towards-agi-and-how-theyre-measuring-it/401/2 "2025-11-26T23:31:37Z")

</div>

> [@ben](#):
>
> Wouldn’t that be extremely stressful? Knowing that the other competitors are a few percentage points away from not only dominating your expertise, but their whole ecosystem?

That’s what competition is for, @ben - keeps you wanting to engage and pedal to the metal. Where the fun without that? 😜

---

<div class="post-metadata">

**Author:** ![Bryan](https://yyz2.discourse-cdn.com/flex034/user_avatar/forum.zurb.com/bryan/32/448_2.png) [@Bryan](https://forum.zurb.com/u/Bryan)\
**Post date:** [November 29, 2025, 11:57pm UTC](https://forum.zurb.com/t/anthropics-latest-the-fight-towards-agi-and-how-theyre-measuring-it/401/3 "2025-11-29T23:57:52Z")

</div>

It’s an interesting time in AI… an all-out arms race, and one that will be won by the company that best addresses user needs… how do you determine that? It always comes down to context and problem.

In the example shown… could be:

- Time to resolution
- Number of steps the user must take
- How many clarifying questions are needed
- Whether the user understands the next action
- User frustration signals (backtracking, repeated complaints)

But those don’t show up necessarily in the model benchmarks. Each model has its own strengths.

---

<div class="post-metadata">

**Author:** ![ben](https://yyz2.discourse-cdn.com/flex034/user_avatar/forum.zurb.com/ben/32/442_2.png) [@ben](https://forum.zurb.com/u/ben)\
**Post date:** [December 4, 2025, 4:58pm UTC](https://forum.zurb.com/t/anthropics-latest-the-fight-towards-agi-and-how-theyre-measuring-it/401/4 "2025-12-04T16:58:06Z")

</div>

As time goes on, I’m seeing a shift in smaller OSS models starting to arise.

Cheaper, faster, open source.

I think it’s gonna take a lot longer before AGI is achieved, so more than likely we’re going to have a world of specialized models that tackle different verticals across the economy.

No single winner…