The money is in implementation, not models.
The biggest AI story for a fund this week was not a model. Anthropic, Blackstone and Hellman & Friedman launched Ode with Anthropic, a $1.5bn firm built on embedding AI engineers inside enterprises. The bet: value sits in the implementation layer, not model access.
Hi folks,
The biggest AI story for a fund this week was not a model. It was the smart money backing the idea that AI's value is in the implementation, not the model.
The one thing
The smart money is betting AI's value is in implementation, not models
- This week Anthropic launched Ode with Anthropic, a firm that embeds AI engineers inside enterprises to build and run their systems. Backers include Apollo, Goldman Sachs, GIC and Sequoia; reported at $1.5bn (TechCrunch).
- The firms funding it are your peers, voting with capital that implementation, not model access, is where the value sits.
- For your own AI, the edge is not the model you licence. It is the governance, auditability and context you own around it.
In the mix
- RBA expects private-credit defaults to keep climbing through 2026 (MPA)
- Why it matters: the central-bank echo of ASIC's scrutiny; a clean, consistent default and valuation methodology is now table stakes for a local credit GP.
- The SEC is pressing fund managers on AI governance through exams, not rules (FinTech Global)
- Why it matters: only 24% of firms have a vendor-AI policy; if you have US LPs, expect the question, and "we use a vendor" is not an answer.
- Meta shipped Muse Spark 1.1, its first paid model, at about a quarter of Fable and GPT pricing (TechTimes)
- Why it matters: a fourth serious entrant in the cheap-and-capable tier; the cost of AI may start falling as competition heats up.
Readings
If you want to double click into things this week I'd recommend:
- The five trends that defined AI engineering in 2026 (Latent Space). This is a recap of the AI Engineer World's Fair. Broadly engineering is 6 months ahead of knowledge work and watching where engineering is going gives a good feel for the industry. The trends:
- Systems over agents. The work has moved from the agent itself to the "harness" around it: the workflow, context, permissions and evals that make it reliable enough to trust.
- Loop engineering. An inner loop where the agent runs, wrapped in an outer loop where a human sets direction and checks the output before it counts.
- Forward-deployed software factories. Engineers embedded inside a business, wiring AI into how the company actually works. This is the Ode model, and ours at Levercon!
- Coding agents replacing the IDE. Tools like Claude Code and Cursor are now the primary interface: you give a goal, the agent iterates across the whole codebase.
- Skills as a discipline. Reusable, plain-language procedures the agent follows and maintains, so a workflow is captured once and reused rather than rebuilt each time.
- Please stop the AI confidence theatre (Elena Verna)
- This is another good read especially if you feel behind on AI.
- Her point: the real value is in the unglamorous iteration, not the demo. "The first prompt is the fun part. The next thousand prompts is where real value is created."
Levercon is the AI team for investment funds. We embed alongside your team to automate workflows, build custom tools, and compound operational gains. Ultimately we help Australian based investment funds leverage AI.
Book a strategy call.
See where AI fits in your fund.
We'll show you where AI lifts efficiency and performance across your organisation.