Google claims Antigravity’s released /boost mode lets developers invoke deeper agent reasoning on demand, providing an explicit quality-versus-latency-and-cost control for coding workflows.

state: expiredheat: lowuncertainty: mediumconvergesscott: mediumcoding-agents agent-harnesses inference-economicsGoogleAntigravity

What is this?

Google Antigravity is presented as Google’s agent-first coding platform, with agents operating across the editor, terminal, and browser and support for orchestrating multiple agents and models. The supplied evidence titles claim a released `/boost` command invokes a deeper multi-agent reasoning pipeline for difficult software-engineering tasks, creating an explicit quality-versus-latency-and-cost control. However, none of the provided web-result snippets directly documents `/boost`, its release status, or its mechanics, so that specific claim remains thinly grounded here.

Why it matters to Scott

Google’s claimed `/boost` control independently productizes Scott’s existing position that coding-agent harnesses should route extra reasoning compute selectively and expose the resulting quality, latency, and token-cost tradeoff. It creates a useful dated-receipts and cross-platform evaluation opportunity alongside Claude Code’s effort controls, but the supplied evidence does not directly verify `/boost` or its mechanics.

dev:concept.task-aware-model-routingip:concept.token-disciplineip:concept.model-plus-harness-benchmark-unitradar:claude-code-effort-controlsradar:concept.coding-agentsradar:concept.agent-harnessesradar:concept.inference-economics

queries asked of Scott's wikis

test-time compute controls in coding agents
quality latency cost tradeoffs in agent harnesses
dynamic reasoning-depth routing for software tasks
multi-agent escalation for difficult coding tasks
user-controlled inference budgets and agent economics
fast-path versus deep-path coding workflows

Measured heat

no measured readings yet — the hourly heat pass fills this in

How the heat travelled

no chain yet — the hourly chain pass fills this in

Evidence (2) — ⭐ canonical anchor

source	object	author	score	comments
🟧 hn	Google Antigravity introduces Boost deep reasoning (/boost)	simonpure	80	57
🟧 echo.other ⭐	The official documentation introduces `/boost` as an on-demand multi-agent reasoning pipeline for difficult software-engineering tasks, incl	Google Antigravity	—	—

Interpretation history

2026-09-04T07:35:12Z
No hands-on measurements or independent implementation evidence arrived within the observation horizon, leaving the established release measurement-bound. The launch episode has faded and can close unless substantive quality, latency, token-use, or cost results emerge.
2026-09-02T07:30:36Z
The larger discussion remains launch amplification and anecdotal product skepticism, not independent evidence about `/boost` outcomes or inference economics. The established release stays measurement-bound pending hands-on comparisons of quality, latency, token use, and cost.
2026-09-01T06:30:32Z
The refreshed discussion remains anecdotal skepticism and launch amplification, with no independent measurements of `/boost` quality, latency, token use, or cost. The established release remains measurement-bound and does not warrant promotion or renewed attention.
2026-09-01T05:40:35Z
The refreshed comments remain repetitive skepticism and launch discussion, adding no independent evidence about `/boost` quality, latency, token use, or cost. The release is established, but the case is now measurement-bound and can cool while awaiting hands-on results.
2026-09-01T04:29:18Z
The refreshed discussion adds only a commenter’s plan and CLI availability detail plus familiar model skepticism; it does not independently validate `/boost` quality or economics. The first-party release remains established, but the case now waits for hands-on measurements or implementations rather than more launch amplification.
2026-09-01T04:28:06Z
grounded: converges/medium — Google’s claimed `/boost` control independently productizes Scott’s existing position that coding-agent harnesses should route extra reasoning compute selective
2026-09-01T04:25:45Z
origin walked (codex/luna, conf 0.97): anchor hn.story.49517537 -> echo.other.866832cc14 by Google Antigravity
2026-09-01T04:24:18Z
case created — A first-party documented control in Google’s coding-agent product creates a distinct, testable workflow and inference-economics episode.

Decision trace

09-04 17:35expireNo hands-on measurements or independent implementation evidence arrived within the observation horizon, leaving the established release measurement-bound. The launch episode has faded and can close un
09-04 17:35alert_silentThe only trigger is staleness; there is no new capability, access, pricing, implementation, or evaluation delta to surface.
09-04 17:35alert_routeThe only trigger is staleness; there is no new capability, access, pricing, implementation, or evaluation delta to surface.
09-02 17:30repriceThe larger discussion remains launch amplification and anecdotal product skepticism, not independent evidence about `/boost` outcomes or inference economics. The established release stays measurement-
09-02 17:30alert_silentThe release was already routed, and the new delta is only engagement growth plus refreshed comments without a material evaluation, implementation, access, pricing, or capability update.
09-02 17:30alert_routeThe release was already routed, and the new delta is only engagement growth plus refreshed comments without a material evaluation, implementation, access, pricing, or capability update.
09-01 16:30repriceThe refreshed discussion remains anecdotal skepticism and launch amplification, with no independent measurements of `/boost` quality, latency, token use, or cost. The established release remains measu
09-01 16:30alert_silentThe first-party release was already routed; refreshed comments add no material capability, access, pricing, implementation, or evaluation result and can wait for hands-on evidence.
09-01 16:30alert_routeThe first-party release was already routed; refreshed comments add no material capability, access, pricing, implementation, or evaluation result and can wait for hands-on evidence.
09-01 16:21sensor_dirtycomment_update
09-01 15:40repriceThe refreshed comments remain repetitive skepticism and launch discussion, adding no independent evidence about `/boost` quality, latency, token use, or cost. The release is established, but the case
09-01 15:40alert_silentThe first-party release was already routed, and the new delta is only refreshed discussion without a material capability, access, pricing, or implementation update.
09-01 15:40alert_routeThe first-party release was already routed, and the new delta is only refreshed discussion without a material capability, access, pricing, or implementation update.
09-01 15:21sensor_dirtycomment_update
09-01 14:29repriceThe refreshed discussion adds only a commenter’s plan and CLI availability detail plus familiar model skepticism; it does not independently validate `/boost` quality or economics. The first-party rele
09-01 14:29alert_silentThe consequential release delta was already routed; higher engagement and unverified comment details are not a material escalation. It can wait for evidence on latency, cost, token use, or coding outc
09-01 14:29alert_routeThe consequential release delta was already routed; higher engagement and unverified comment details are not a material escalation. It can wait for evidence on latency, cost, token use, or coding outc
09-01 14:28alert_shadowGoogle’s first-party documentation establishes that `/boost` is available as an explicit escalation path using strategy formation, specialized parallel agents, and independent verification. This is a
09-01 14:28alert_routeGoogle’s first-party documentation establishes that `/boost` is available as an explicit escalation path using strategy formation, specialized parallel agents, and independent verification. This is a
09-01 14:28groundGoogle’s claimed `/boost` control independently productizes Scott’s existing position that coding-agent harnesses should route extra reasoning compute selectively and expose the resulting quality, lat
09-01 14:25promote_anchororigin walk conf 0.97
09-01 14:24createA first-party documented control in Google’s coding-agent product creates a distinct, testable workflow and inference-economics episode.