Co-Founder/CEO @PeakMetricsApp
Los Angeles, CA
Joined January 2009
- Tweets53
- Following2.6K
- Followers1.7K
- Likes2.3K
Pinned Tweet
1/ I've spent my career building technology and leading digital marketing teams. The techniques we used to reach audiences and grow businesses are being used to manipulate people. That realization led me to co-found @peakmetricsapp.
Nick Loui retweeted
While evaluating Qwen3.8-Flash-Next in a controlled sandbox environment, Cohere's research team observed something unexpected.
Asked to generate a defensive cybersecurity detection rule, the model instead inventoried aspects of the machine it was running on and attempted to upload a probe file to an Alibaba Cloud storage bucket in Beijing. Notably, the model's reasoning trace stated that it would not invoke any tools before taking these actions.
Cohere's sandbox blocked the attempt, and no external communication succeeded. We have not determined the root cause, and we are not asserting malicious intent.
But the incident raises important questions for organizations deploying agentic AI systems: Can models take actions that differ from their stated reasoning, and how should security teams evaluate trust, provenance, and behavior in open-weight models?
We are publishing our findings because organizations making model adoption decisions deserve as much information as possible. We hope this contributes to greater transparency, stronger security practices, and continued collaboration across the AI ecosystem.
Read the full report: cohere.com/blog/concerning-b…
Those call for different actions: make the improvement easier to discover, fix the usability problem, or investigate how the claim arose.
An old review may describe a problem you’ve already fixed. A recurring complaint may point to friction customers still experience. A claim without supporting evidence needs verification before anyone reacts.
If an AI answer calls your product “hard to use,” what do you fix?
First, inspect the answer’s sources. Then check whether the criticism holds up.
We checked the claims that the Yandex strikes silenced Russian influence operations.
Our team at @peakmetricsapp looked at suspected bot activity across 20 topic areas on October 11, comparing it with nine previous Sundays. That matters because weekend activity is often lower, even without an outage.
18 of the 20 stayed within their usual Sunday range. Two were below it. We didn’t see the broad collapse the headlines suggested in our sample.
The Cloudflare chart behind the story measures automated web requests from Yandex’s network, which can include search crawlers. It doesn’t tell us which social accounts are run by Russian operators, or whether those accounts stopped posting.
Before declaring an influence operation offline, we need evidence from the accounts and conversations it was driving. A sharp drop in web traffic alone doesn’t get us there.
radar.cloudflare.com/bots/as…
At @peakmetricsapp, @typesafeai’s Jev showed ~82% lower cost per doc and ~47% lower median call latency vs our baseline. A blind LLM judge preferred Jev in 69% of 200 disagreements. Workloads and batch sizes differed, so these are directional, but very promising results.
Legora's CRO says they hire roughly two post-sale people for every one pre-sale. As AI capabilities improve, the human experience around them can become a bigger differentiator. Having someone who understands your work and helps you get it right is still worth a lot.
NEW: How @WeAreLegora scaled from $3.5M to $200M in revenue 📈
They were the fastest company ever to hit $100M with a direct sales motion, in just 18 months. Then they DOUBLED it to $200M ARR in less than 6 months 🤯
Patrick Forquer (CRO at Legora) joins GTMnow to break down how they scaled that fast: pilot design, why implementation predicts NRR, screening AEs for AI fluency, and much more.
Pat joined Legora at around $3.5M after seven years at Braze, from the early days through the IPO. He built out the US presence, starting from the first “office” aka his Brooklyn apartment, where his son was on snack duty.
Highlights:
01:40 The three criteria for assessing joining a company: founder, product, investors
06:40 Advice for operators weighing the jump to an AI-native company
09:10 The early days: first US pipeline, and an office in a Brooklyn apartment
12:00 From $3.5M to over $100M: the fastest-ever scaling
13:10 The two biggest GTM unlocks: customer obsession and the pilot experience
15:30 Hiring 2-to-1 post-sale, and forward-deploy structure
18:30 Why time-to-value is the no.1 leading indicator of revenue
21:00 How to think about forward-deployed engineers for your own product
24:00 The pilot design: let customers keep what they build
28:00 The "LFG" culture (it doesn't stand for what you think...!), and shipping major features every few weeks
30:00 The near-death stuff: a full sales tech-stack overhaul mid-scale
32:30 The Jude Law brand campaign, and whether it was worth it
36:00 The Aaron Judge and Yankees partnership
38:30 Measuring brand ROI: pipeline and a brand index
47:00 What a great AE at Legora does differently
50:00 Screening for AI fluency: 101 vs. 301-level
52:00 Handling competitive objections, and the "8 Mile" framework
On screen: Patrick Forquer x @sophiebuona
Full interview available here on X, on YT, Spotify or Apple.
8/ That work now extends to AI answers. Last month, we launched AI Perceptions to track how brands appear in AI responses alongside news and social narratives.
peakmetrics.com/insights/pea…
Nick Loui retweeted
We’re excited to share that PeakMetrics has closed our Series A led by @monetaventures, bringing total capital raised to $16.3M.
Narratives now move faster than ever, and they’re engineered. We’re building the intelligence to see through it.
Read more: peakmetrics.com/insights/pea…
