zeroshot v0.10.0

insights · token efficiency, not token maxing

Same battery.
Different mileage.

Everyone’s capped at the same token budget. Same models, same limits, same ceiling. The question is who gets the most done with it — and ZeroShot gives you the receipts.

zeroshot insights · ~/api · mon 9:14am
❯zeroshot insights --month
⏺usage · this month · your team
└tokens saved: 2.4M · tool calls intercepted: 18,402
└efficiency score: 88 — was 76. still climbing.
└output per token: +30% mom · corrections: −42%
⚠flagged: ~38% of last week’s tool calls & tokens recoverable — run the scan ↓
❯

[1] · are you getting better?

Watch your team get better at this.

Is your codebase getting easier to work in? Are your agents getting more efficient? ZeroShot reads your history and tells you straight — fewer corrections, less friction, and more shipped per token, month over month.

+30%
more efficient — output per token, month over month
−42%
corrections — less steering, fewer re-litigated reviews
76 → 88
efficiency score, and still climbing

[2] · we find the waste

See exactly where your tokens
& tool calls leak.

ZeroShot scans your agent sessions, flags the tool-call and token inefficiency — re-derived context, rebuilt UIs, re-litigated reviews — and suggests the exact skill that fixes each, with the savings attached.

inefficiency scan · last 7 days
❯zeroshot scan --last 7d
⏺scan complete · ~38% of last week’s tool calls & tokens recoverable
─── flagged leaks ───
✗Re-deriving context already in a teammate’s sessionTool calls
└14 sessions started cold instead of resuming an existing one.
└wasted 4,200/wk → fix: bb agent-sessions resume → saves ~3,800
✗Re-reading the same files every runTool calls
└The same 8 files grepped & read across 12 sessions.
└wasted 3,400/wk → fix: /bb-analyze → saves ~3,000
✗Rebuilding UIs before the design settlesTokens
└The dashboard was rebuilt 3× before a layout stuck.
└wasted 210k/wk → fix: prototype in ZeroShot → saves ~150k
✗Re-litigating the same review conventionsTokens
└11 PRs re-flagged the same 3 conventions by hand.
└wasted 140k/wk → fix: /bb-review → saves ~120k
✗Building the wrong thing firstTool calls
└Specs shipped without the customer context to scope them.
└wasted 2,100/wk → fix: /bb-specify → saves ~1,700
─── recent sessions · efficiency score ───
auth/oauth-rewrite· nikhil91✓ resumed prior session
payments/stripe-webhook· maya68⚠ rebuilt UI 2× — try prototypes
codex/llm-data-source· spencer52⚠ could’ve used /bb-analyze
#Estimated from your sessions. Apply a suggested skill and ZeroShot tracks the actual savings.

[3] · the leaderboard

Who’s a surgeon,
who’s a shotgun.

ZeroShot scores your team on output per token. Spending the most tokens isn’t the flex — getting the most done with them is.

TOKEN EFFICIENCY · THIS WEEK
PRs shipped / Mtok
#1nikhilSurgeon98
9.8M tokens · 14 PRs
#2maya86
11.2M tokens · 12 PRs
#3daniel71
14.1M tokens · 11 PRs
#4spencerToken maxer52
21.6M tokens · 9 PRs
note ─ spencer burned 2× the tokens for fewer PRs. nikhil ran surgical. same battery, different mileage.

[4] · prove it to your CTO

See exactly what you saved.

Tool calls intercepted, tokens saved, broken down by team. The number a CTO needs to sign off — and the proof your engineers are getting twice as far on the same budget.

usage · last 6 weeks · illustrative
2.4M
tokens saved this month
18,402
tool calls intercepted
Savings by team
Platform1.1M
Growth740k
Success520k
# Coverage: 100% of your sessions.

[5] · same feature. half the tokens.

You’re not blocked on talent.
You’re blocked on
using your tokens better.

When your agent has your team’s skills, it gets the build right the first time — no re-deriving context, no re-litigating conventions, no burning a review cycle on the wrong thing.

tokens to ship one feature · illustrative

Same feature, same agent. The difference is whether your team's skills intercept each step — so the agent gets it right the first time instead of iterating.

~52%
fewer tokens per feature
= ship 2× more on the same budget
Raw-dogging it≈ 1.2M tokens
↻ Re-derives context, re-litigates conventions, retries the wrong build
With ZeroShot CLI≈ 580k tokens
skills intercept:/bb-specify/bb-plan/bb-review/trust-but-verify
# Illustrative estimate. Actual savings depend on your repo, your skills, and how often your team is repeating work the agent could have inherited.

[6] · the full picture

Everything your agent sessions
can tell you.

One read of your sessions and the whole picture surfaces — what you’re working on, where the tokens go, and how it’s all trending.

→ what kind of work

Token spend split across research, features, bug fixes, and refactors — see where your team's effort actually goes.

→ where tokens go

Spend broken down by repo, person, and PR, with a 30-day trend and month-end projection.

→ efficiency score

A 0–100 score from token windows, tool pressure, signal coverage, and repair / scope drag.

→ speed & velocity

Time saved, shipping speed, and tokens-per-ship — before and after your skills kicked in.

→ coverage

Which codebase areas your skills actually cover — skill-system, docs, tests — with a confidence read on each.

→ ranks

How you stack up against your team and the global ZeroShot cohort on efficiency and output.

→ extraction & signals

Every friction event, signal type, and piece of evidence ZeroShot pulled from your sessions.

→ cost

Token reduction: current vs. projected spend, with savings attributed to the work that's actually reducible.

→ top friction sessions

The sessions that took the most steering, ranked — the back-and-forth quietly draining your budget.

the receipts · the bill came due

"The conversations are never about capability now. Now they’re about, ‘hey, we’re spending so much.’"

— Alexander Embiricos, Head of Enterprise, OpenAI · via TechCrunch →

"One of my engineers spent $40,000 on tokens last month, and I genuinely don’t know whether I should stop him."

— Vitaly Gordon, CEO, Faros AI · via TechCrunch →

"I'm blocked from using agent tools the second I hit my budget. ZeroShot unblocks me — same cap, twice the output. That's the whole pitch."

— Nikhil Shinday, CTO

Everyone is asking the question. ZeroShot is the answer with receipts attached — not "trust me," but tokens saved, tool calls intercepted, per team, per week.

❯ ok. give it to me.⏺ two ways in:
Download for macOSv0.10.0 · 13.4 MB · free, no account
Book a demo for your team →SOC 2 · can run locally