Rosenblatt — The Goldilocks Zone with Open Source Models

The Strategy

The Goldilocks Zone

Too Hot

All-frontier, all the time

Flagship intelligence on every call — and unit economics that collapse the moment you build loops or scale up delegation.

Just Right

Multi-family, OSS models

Ensembles of OSS models hit frontier-level quality at loop-friendly prices — with true frontier models held back for surgical moments.

Too Cold

Self-hosting for everything

Hits an intelligence ceiling. The model's blind spots become your blind spots — and a single point of failure.

Part 1

The Harness

The Frontier Rug Pull

April 7th: Anthropic Bans Us

Without warning or explanation, all team member accounts are banned.

No reason was provided by Anthropic for the bans. Many appeals have been filed with no response and we have failed to self diagnose any policy violations.

Luckily, we were already heavy GitHub CoPilot users

In April, GitHub Copilot use was extremely subsidized (+90%). The ban pushed us deeper into our coding harness and further from the Anthropic ecosystem.

2 months later, the ban was randomly lifted on all but one account.

As suddenly as we lost access we regained it. We still have no response regarding the remaining account, but we couldn't care less.

The April Off Site

Michiana, Michigan — Hacker Beach House

Paused all client work for 1 week to focus on R&D and team bonding.

Explored OpenClaw, Paperclip, Hermes, and similar frameworks.

Walked away with valuable insights and new plans for 2026.

Harnesses

An Unlikely Combo: Hermes + Copilot SDK

Turn Hermes into our digital twins, one per team member.

Explore tokenmaxing on the heavily subsidized Copilot plans.

Leverage GitHub Copilot BYOK for full provider freedom.

Why Hermes

Like OpenClaw with Self-Improvement

OpenClaw was an amazing proof of concept, but far from production ready.

Insecure by default, unmoderated plugin library, and poor observability has kept it from being adopted by enterprise.

Hermes is a master of skill management and self improvement.

We don't trust Chinese-hosted endpoints, so we don't use them. Distrusting a host doesn't mean discarding the model.

With Hermes, coalescing agentic coding and agentic servicing is possible.

Many teams have completely different harnesses for coding and servicing, but Hermes unifies them under a single framework.

Part 2

The Infrastructure

The Trust Model

Chinese Models, American Providers

Open weight models are simple, verifiable math formulas — not software.

Anyone can download the weights and run them anywhere. Nothing about inference has to touch the lab that trained the model.

The provider is the risk surface — not the model.

We don't trust Chinese-hosted endpoints, so we don't use them. Distrusting a host doesn't mean discarding the model.

US providers with zero data retention — no exceptions.

Every token we run goes through American providers with ZDR policies: prompts and outputs aren't stored, logged, or used for training.

The Providers

Two Providers, No Usage Limits, No Outages

Primary

From the creators of PyTorch, NVIDIA-backed, and publicly praised by Jensen Huang. Raised a $1.5B Series D in July 2026 at a $17.5B valuation on $1B+ annualized revenue.

Our default endpoint for every family we run.

Fallback

Newer, and usually beats Fireworks on price. $107M Series B raised in May 2026, new Toronto data center opened in August 2026. Latency is less consistent and we've hit the occasional small bug.

Crucially: it hosts the same models we already run.

Same weights on both sides means failover is a routing change — not a prompt-engineering project.

Failover in Action

IRL: the GLM-5.2 incident

01

Fireworks wobbles

Fireworks starts having trouble serving GLM-5.2, our most popular endpoint, and a 90 minute outage ensues.

02

Flip the switch

Traffic reroutes to DeepInfra, which hosts the same GLM-5.2. Same weights, same prompts, same harness.

03

Nothing else changes

Work carries on without incident — near-100% behavioral consistency, because we never changed the model.

Redundancy across providers. Consistency within the model. That's the whole design.

Subscribe, Don't Meter

Keep the frontier — Subscribe, don't meter

The Key

Run Anthropic's flagship models exclusively through flat $25–$125/m subscription tiers on a business team agreement, never API rates. Use Opus & Fable as surgeons' tools; break glass in case of emergency. Extremely useful — deliberately rare.

DO

Reserve it for absolute-flagship reasoning.
Leverage it for complex debugging and long-form writing.
Make it your default for cheap deep research.

DON'T

Build loops on frontier models.
Assume planning must start at the absolute frontier.
Pay API rates for capacity a subscription already covers.

Part 3

The Models

The Shift

The ground shifted — We rebuilt around it.

100%

of our day-to-day inference now runs on open-weight models.

14x

saved when compared to Anthropic models at API rates.

6+

production model families (e.g. DeepSeek, GLM, Qwen, Kimi, Nvidia, MiniMax, etc.).

The frontier labs still own the very top end — the game is deciding where those tokens are actually worth it.

Auxilary Models

Auxiliary Models: delegate case by case

flowchart LR T["Task arrives
prompt + screenshot"] Q["Qwen 3.8 Max (Vision)
describes the image in detail"] G["GLM-5.2 (Default)
text-only default"] A["DeepSeek Flash (Worker)
executes the task"] T --> G T --> Q Q --> G G <--> A classDef box fill:#141414,stroke:#2a2a2a,stroke-width:1.4px,color:#ece8e3; classDef cold fill:#141414,stroke:#7fa5c4,stroke-width:1.6px,color:#7fa5c4; classDef gold fill:#141414,stroke:#c676e4,stroke-width:1.6px,color:#c676e4; classDef out fill:transparent,stroke:#c2825c,stroke-width:1.6px,color:#ece8e3; classDef green fill:#141414,stroke:#5fc770,stroke-width:1.6px,color:#5fc770; class T box class Q cold class G gold class A green

Build Hermes into your digital twin — make it think and work like you.

Tokenmax on the heavily subsidized Copilot plans.

Swap per gap without touching the harness.

Mixture of Agents

MoA: Don't bet on a single model family

flowchart LR P["Prompt"] D["DeepSeek V4 Pro
reference model"] G["GLM 5.2
reference model"] K["Kimi K3
reference model"] A["Qwen 3.8 Max
aggregator model"] O["Frontier-level
answer"] P --> D P --> G P --> K D --> A G --> A K --> A A ==> O classDef box fill:#141414,stroke:#2a2a2a,stroke-width:1.4px,color:#ece8e3; classDef cold fill:#141414,stroke:#7fa5c4,stroke-width:1.6px,color:#7fa5c4; classDef gold fill:#141414,stroke:#c676e4,stroke-width:1.6px,color:#c676e4; classDef green fill:#141414,stroke:#5fc770,stroke-width:1.6px,color:#5fc770; class P box class D,G,K cold class O green class A gold

Share the prompt across multiple models.

Aggregate each model response.

Return a final answer with frontier quality.

Split The Work

Planner → Workers → Reviewers

Stop asking one model to do everything. Route each phase to the cheapest model that can own it.

Planner 30%

Use a top OSS model or MoA to turn the task into an explicit plan before any execution starts. This is where quality gets decided.

Workers 60%

The plan is delegated to one or more DeepSeek Flash-class models that grind through execution. This is where loop volume lives.

Reviewers 10%

High-intelligence, but lower context window models grade the output — ideally several independent reviewers across several families.

The Playbook

The Goldilocks Playbook — and what's next

01
One harness for code and service — Hermes is uniquely comprehensive.
02
Open weights on trusted hosts — Fireworks AI primary + DeepInfra fallback.
03
Primary + fallback, same models — Fireworks first, DeepInfra right behind it.
04
Keep the frontier subscription — use for hard calls and edge cases, not loops.
05
Diversify across model families — balance your quality and spend.

Next Step

Trial Fireworks Nexus as a general harness endpoint

Fireworks new model router offers a new savings strategy to route requests efficiently based on your preferences (balanced, max-savings, and max-intelligence). Fireworks reports 54% overall AI cost savings with their early access customers.

Not too hot, not too cold — frontier results at open-source prices.

Finding the Goldilocks Zone with Open Source Agents