Home/Infra/@testmachine_ai
70
Score · promising
Scored 2mo ago · may be stale

@testmachine_ai

TestMachine

TestMachine is an AI-powered smart contract security platform featuring Azimuth, an autonomous attack agent that tests contracts on-fork to find exploitable vulnerabilities. The platform reports 86.3% detection rate on EVMBench with ~2.7% false positives, has caught multiple critical vulnerabilities in live protocols worth millions, and recently disclosed a Ledger Ethereum app vulnerability affecting all users. Founded Dec 2022, ~2.4k followers, backed by undisclosed investors, the team has demonstrated real-world impact including confirmed bounties and catching a $75M Tectonic exploit pattern.

AI Analysispromising

Confidence
82%

TestMachine is an AI-powered smart contract security platform featuring Azimuth, an autonomous attack agent that tests contracts on-fork to find exploitable vulnerabilities.

The platform reports 86.3% detection rate on EVMBench with ~2.7% false positives, has caught multiple critical vulnerabilities in live protocols worth millions, and recently disclosed a Ledger Ethereum app vulnerability affecting all users.

Founded Dec 2022, ~2.4k followers, backed by undisclosed investors, the team has demonstrated real-world impact including confirmed bounties and catching a $75M Tectonic exploit pattern.

Green flags: Concrete technical performance metrics (86.3% detection on EVMBench, 2.7% false positive rate) · Multiple confirmed critical vulnerability discoveries in live protocols with bounties paid · Responsible disclosure track record (Ledger vulnerability disclosure, Tectonic analysis) · Early-stage footprint (<2.5k followers, account <2 years old) with working product · Integration with established platforms (Sherlock Audit Engine participation) · Real testimonials from Coinbase security engineer and independent auditors

Red flags: Team members not named or prominently featured on website · Investor backing mentioned ('Backed by the Best') but backers not disclosed · Business model and token strategy unclear (no token mentioned)

Token
No · pre-launch
Chain
Ethereum
Stage
mainnet+live
Category
smart contract security

Recent tweetsSee all on 𝕏 →

What can a better observability layer for your AI agents do for your workflow? TestMachine lets you enforce policy and guidance on your agents and finds new policies to cover risks you didn't anticipate. See what TestMachine can do for you today. https://t.co/SdyWHodr2P
1w ago♥ 15💬 13🔁 0
We’re excited for everyone to try out TestMachine, a trust and observability layer for AI agents. TestMachine records connected tool calls, checks them against the rules your team writes, and blocks calls that break them before execution. What makes us different from other tools? Follow the thread: Check out our blog: https://t.co/50C08iRQi5
1w ago♥ 22💬 16🔁 6
We're excited to announce the launch of TestMachine. An MCP gateway layer that gives you powerful observability and monitoring of your AI agents. Does your team have many agents without much visibility between people? TestMachine's observability layer allows your team to maximize collaboration and adoption of AI agents. TestMachine observes agent activity and, with our state of the art data factory, can create complex situations to test how your agents react, so we can build policies with you to protect you against unauthorized actions. Check out our blog and try it out for yourself today! https://t.co/50C08iRQi5
1w ago♥ 0💬 0🔁 0
288 AI incidents were reported between January 1 and October 22 this year. In February, an AI coding agent deleted a course platform's production database, along with its backups. The agent had permission to do it. Nobody had written the rule that should have applied. Teams are giving agents real work before they know which rules those agents need. So we built TestMachine to help teams build and enforce those policies. It shows your team what your agents actually do. You write rules in plain English, and TestMachine checks each connected action against them before it runs, allowing or blocking it. https://t.co/50C08iRQi5
1w ago♥ 0💬 2🔁 0
Lots of companies are adopting AI agents into their everyday workflow without knowing all the risks. You could set policies and still be exposed to unauthorized actions you never thought of. For instance, prompt injection is a major attack that agents are susceptible to. Say an agent reads and replies to your emails. An attacker can hide instructions in a message, and the agent may follow them without you noticing. This is just one way agents can take actions you never expected. Your guardrails and policies may not be enough. Do you think you are prepared and covered against all forms of AI agent risks?
1w ago♥ 1💬 1🔁 0

Signal Timeline

EC
@econoar followed
BFirst discovered·2mo ago

Score breakdown0–100

🎯Scout conviction
+23.8 / 35
📚Scout consensus
0 / 10
🪪Profile & earliness
+14 / 20
✍️Substance
+16 / 20
🤖AI verdict
+22.3 / 30
⚠️Penalties
-6 / 40
70
Below threshold (74)
Watching for additional signals.
Followers
2.4K
Account age
3.9y
Scouts
1
First seen
2mo ago