Synth GTM brain

Synth helps AI engineering teams optimize coding agent prompts and context against a chosen harness, model and task dataset.

category

Coding agent optimization

market profile

  • text: Multi-step agent runs create long trajectories that are difficult to inspect and optimize by hand.
  • text: Synth's current positioning focuses on coding agent context and prompt optimization; its earlier launch emphasized tracing failures and fine-tuning models.

personas

  • title: Coding agent engineer
    pains:
    • Finding the cause of failures in long, multi-step agent traces
    • Improving prompts and context against representative coding tasks
  • title: AI research engineering lead
    pains:
    • Comparing harness and model choices consistently on real workloads
    • Improving task completion without slow manual prompt experimentation
  • title: Developer tools product lead
    pains:
    • Coding agents fail on lengthy, ambiguous tasks
    • Needs an evaluation set that reflects the product's user workflows

icp

  • text: Teams building coding agents that have a task dataset, an agent harness and a model to optimize; target employee count is not public.
  • text: AI research engineers operating long-running coding workflows are a close fit; Synth's documented stack includes local and hosted tools.

objections

  • objection: Will optimization improve our specific coding workflows?
    response: Benchmark on representative held-out tasks against the existing harness, model and prompts.
  • objection: Will our model, environment and data be supported?
    response: Confirm harness support, execution environment and code handling directly before running proprietary tasks.
  • objection: Could we optimize manually or with open tooling?
    response: Compare the engineering time and measured task success of each approach on the same dataset.

competitors

  • name: Mentlio
    why: Listed as a nearby company with overlapping coding, prompt and context optimization terms; precise feature overlap requires validation.
  • name: Jcode
    why: Listed as a nearby coding agent and agent harness company; overlap is in the coding agent workflow.

influencers

  • name: Josh Purtell
    url: https://www.ycombinator.com/companies/synth-3
    why: Founder and CEO with experience researching agent optimization.

keywords

  • coding agent prompt optimization
  • agent context optimization
  • coding agent evaluation
  • long horizon agent benchmarks
  • agent failure analysis
  • agent harness optimization

signals

  • text: A team is assembling coding agent benchmarks and tuning prompts or context for a specific harness.
  • text: An AI engineering team is struggling to diagnose long agent trajectories and recurring failures.

stack

  • GEPA
  • StackEval
  • synth-dev

pricing

model: unknown

motion

unknown

channels

  • YC launch
  • LinkedIn
  • Documentation

confidence

low

Founders

  • name: Josh Purtell
    title: Founder/CEO
    role: CEO
    bio: Josh is the founder and CEO of Synth. Before Synth, Josh was employee #3 and research scientist at Basis, which uses AI agents to automate accounting tasks. Previously, he built out the ML core of a Cyber Threat Intelligence SaaS product at Halborn, a cyber security startup that acqui-hired his startup Politeia. He founded Politeia while wrapping up a Math undergrad at Yale, where he did time-series ml research for fun.

Back to YC companies