AI-OS Studio
TC
CG
CodeGraft SystemsEvaluator Edition

Grant & Benchmark Evaluation Sandbox

Request Prime Tier
Trial Limitations Enforced: Capped at 25 context slices per session. Single-module boundary scope active (multi-repo graph mesh locked to enterprise). Black-box AST execution active.
Baseline Tokens
285
Status quo full-file dump
Surgical Tokens
143
Isolated symbol + contracts
Token Reduction
50% SAVINGS
Consistent multi-turn reduction
Local Latency
6.4 ms
Zero cloud API overhead

Select Benchmark Workload or Enter Custom Code

Choose a preset module to test surgical boundary extraction

Target Symbol:
STATUS QUO : UNFILTERED PROMPT
285 tokens (Raw dump)
import { StripeClient } from "./clients/stripeClient.js";
import { AuditLogger, WebhookDispatcher } from "./telemetry.js";
import { MetricsCollector } from "./monitoring.js";

export interface Receipt {
  chargeId: string;
  status: "PAID" | "FAILED";
  amountCents: number;
}

export class PaymentService {
  private stripe = new StripeClient();
  private audit = new AuditLogger();
  private metrics = new MetricsCollector();

  // 120 lines of webhook queues, retry listeners, telemetry tables ...
  public async handleWebhookRetry(eventId: string): Promise<void> {
    this.metrics.increment("webhook.retry");
    await this.audit.log("WEBHOOK_RETRY", { eventId });
  }

  // Target Symbol: capturePayment
  public async capturePayment(chargeId: string): Promise<Receipt> {
    const charge = await this.stripe.charges.capture(chargeId);
    return {
      chargeId: charge.id,
      status: "PAID",
      amountCents: charge.amount
    };
  }

  // 140 lines of currency conversions, tax tables, and compliance audits ...
  public convertCurrency(cents: number, fxRate: number): number {
    return Math.round(cents * fxRate);
  }
}
Dumps full file with unneeded handlers, inflating tokens and risking hallucination.
CODEGRAFT : SURGICAL CONTEXT SLICE
143 tokens (50% SAVINGS)
// [CodeGraft Systems : Evaluator Edition | Active Closed Alpha Trial]
// Target: capturePayment | Latency: 5.92ms | Savings: 50%
import { StripeClient } from "./clients/stripeClient.js";
import { AuditLogger, WebhookDispatcher } from "./telemetry.js";
import { MetricsCollector } from "./monitoring.js";
export interface Receipt {
  chargeId: string;
  status: "PAID" | "FAILED";
  amountCents: number;
}

// --- SURGICAL SYMBOL DEFINITION ---
public async capturePayment(chargeId: string): Promise<Receipt> {
    const charge = await this.stripe.charges.capture(chargeId);
    return {
      chargeId: charge.id,
      status: "PAID",
      amountCents: charge.amount
    };
  }
Surgically bound imports & interfaces. Zero noise reaching the model prompt.6.4 ms

For Google Antigravity, Anthropic & OpenAI Reviewers

3 to 6 Month Benchmark Grant

We are conducting parallel multi-turn benchmark evaluations across 50+ enterprise repositories (100k+ to 1M+ LOC) comparing agent completion rates with and without CodeGraft. To lift trial rate limits and run high-throughput parallel benchmark jobs on your flagship models:

Direct Contact:timir.c@olivescale.com•CLI Evaluator:node bin/codegraft-eval.js