A common approach to automating Amazon shopping or similar complex websites is to reach for large cloud models (often vision-capable). I wanted to test a contradiction: can a ~3B parameter local LLM model complete the flow using only structural page data (DOM) plus deterministic assertions?

This post summarizes four runs of the same task (search → first product → add to cart → checkout on Amazon). The key comparison is Demo 0 (cloud baseline) vs Demo 3 (local autonomy); Demos 1–2 are intermediate controls.

More technical detail (architecture, code excerpts, additional log snippets):

https://www.sentienceapi.com/blog/verification-layer-amazon-...

Demo 0 vs Demo 3:

Demo 0 (cloud, GLM‑4.6 + structured snapshots) success: 1/1 run tokens: 19,956 (~43% reduction vs ~35k estimate) time: ~60,000ms cost: cloud API (varies) vision: not required

Demo 3 (local, DeepSeek R1 planner + Qwen ~3B executor) success: 7/7 steps (re-run) tokens: 11,114 time: 405,740ms cost: $0.00 incremental (local inference) vision: not required

Latency note: the local stack is slower end-to-end here largely because inference runs on local hardware (Mac Studio with M4); the cloud baseline benefits from hosted inference, but has per-token API cost.

Architecture

This worked because we changed the control plane and added a verification loop.

1) Constrain what the model sees (DOM pruning). We don’t feed the entire DOM or screenshots. We collect raw elements, then run a WASM pass to produce a compact “semantic snapshot” (roles/text/geometry) and prune the rest (often on the order of ~95% of nodes).

2) Split reasoning from acting (planner vs executor).

Planner (reasoning): DeepSeek R1 (local) generates step intent + what must be true afterward. Executor (action): Qwen ~3B (local) selects concrete DOM actions like CLICK(id) / TYPE(text). 3) Gate every step with Jest‑style verification. After each action, we assert state changes (URL changed, element exists/doesn’t exist, modal/drawer appeared). If a required assertion fails, the step fails with artifacts and bounded retries.

Minimal shape:

ok = await runtime.check( exists("role=textbox"), label="search_box_visible", required=True, ).eventually(timeout_s=10.0, poll_s=0.25, max_snapshot_attempts=3)

What changed between “agents that look smart” and agents that work Two examples from the logs:

Deterministic override to enforce “first result” intent: “Executor decision … [override] first_product_link -> CLICK(1022)”

Drawer handling that verifies and forces the correct branch: “result: PASS | add_to_cart_verified_after_drawer”

The important point is that these are not post‑hoc analytics. They are inline gates: the system either proves it made progress or it stops and recovers.

Takeaway If you’re trying to make browser agents reliable, the highest‑leverage move isn’t a bigger model. It’s constraining the state space and making success/failure explicit with per-step assertions.

Reliability in agents comes from verification (assertions on structured snapshots), not just scaling model size.

Welcome to the Quake Brutalist Game Jam

Testers wanted: VM-curator v0.2.0 with GPU pass-through support

The Burning of Minneapolis

Amazon bungles Wednesday layoff plan with misfired internal email

NBA Grizzlies Owner Robert Pera Linked to Russia's "Crimes Against Humanity"

A verification layer for browser agents: Amazon case study

Show HN: A free online video compression tool for instant compression

Where China built on coal, India is building on sun

What 8 Weeks of Cold Exposure Taught Me

Artemis 2 launch: Livestream info, launch window details

FASB 56 and the Authority of the DNI to Waive SEC Financial Reporting

The Silicon Valley Canon: On the Paideía of the American Tech Elite (2024)

Spree: Open-source eCommerce platform Built using Ruby on Rails

Is coding dead because AI has taken over it?

Kimi AI K2.5 Model Introduction [video]

Amazon inadvertently announces cloud unit layoffs in email to employees

Internet Movie Cars Database

Firefox Split View is ready for testing

How to Win Titular Metagames

The Rise of Chinese Memory [video]

The Thing About Moltbot (Clawdbot) That Nobody Wants to Admit

Super Monkey Ball ported to a website

Human-centred cybersecurity: the case of the Florida water plant hack

Matrix on Cloudflare Workers; what could go wrong?

AI found 12 vulnerabilities in OpenSSL

Maine's Immigrant Students Stay Home as ICE Operation Ramps Up

Why are we still so afraid of using the grumpy old period?

Washington State Bill Seeks to Add Firearms Detection to 3D Printers

Should professors be forced to retire?

I wrote a TTF-to-Texture function for SDL2

Ask HN: What happens when AI coding crosses the human dev threshold