The agent promised 2 order updates. Made 1. Reported 2.
All 4 baseline τ-bench conversations completed both updates. A Kashikoi simulation exposed the omission.
The Conversation is the Test: blog.getkashikoi.com/2026/10… (1/12)
Tim Michaud retweeted
Run, do not walk, from any cloud provider who is throwing public shade at Railway (or any other provider). The better your security team is, the more internal pressure you have not to mouth off about security in Q4 2026.
Tim Michaud retweeted
I’m #hiring an individual contributor for a fully remote, global role at the intersection of vulnerability research, exploit development, and ML/AI — with a focus on fine-tuning open-weight #LLMs. 🧠
I’m not looking for an “LLM whisperer” or an “LLM pilot.” 🚫
I’m looking for someone who deeply understands post-training, data, evaluation, and how to make models reliable in real-world environments. 🔐
The application link is in the first comment. 🌍
#Hiring #LLM #AI #ML #FineTuning #CyberSecurity #llmwhisperer #llmpilot
Tim Michaud retweeted
Over time, I have developed a bit of a reputation for saying no to external work for my teams. At least that's how people describe it. But I realized what I actually do is shape the work, often before it even arrives.
andywgrant.substack.com/p/it…
Tim Michaud retweeted
Rich Sutton's Bitter Lesson Isn't the Whole Story
The tech world has a favorite essay. Walk into any AI strategy meeting, and someone will inevitably reference it to justify why they need more compute, more data.
But here's how to spot a pseudo-intellectual: they never mention Sutton's equally important essay on "Verification."
blog.getkashikoi.com/2025/11…
Tim Michaud retweeted
At long last - Phrack 72 has been released online for your reading pleasure!
Check it out: phrack.org/
Tim Michaud retweeted
Kashikoi is a simulation engine to benchmark GenAI Agents. They generate CPU-friendly world models that autonomously interview agents and generate deep behavioral assessments.
Congrats on the launch, @aaksham and @TimGMichaud!
ycombinator.com/launches/Nfa…
Tim Michaud retweeted
The LLM Data Company creates tooling to write, version, and execute evals for models and agents. They help you measure performance and define rewards for reinforcement learning.
Congrats on the launch, @bertgodel, @thegavinbains, and @joeybesgen!
ycombinator.com/launches/Nai…
Tim Michaud retweeted
cubic (@cubic_dev_) is Cursor for code review.
An AI-native platform that helps teams like Cal.com, n8n, and the Linux Foundation ship code 28% faster.
Congrats on the launch, @NeuralPaul and @allisonee2!
ycombinator.com/launches/NYz…
Tim Michaud retweeted
Well, I’m officially homeless. If anyone can help out with a job I’d be forever in your debt
Tim Michaud retweeted
A fish called Gemmy and how it saved @cmwdotme. A story I’ve wanted to tell for a long time. semafor.com/article/01/10/20…
Tim Michaud retweeted
My team at Apple is hiring for a role focused on compiler tools and infrastructure. If you're interested, check out this role: jobs.apple.com/en-us/details…
Tim Michaud retweeted
Unpopular opinion: companies doing an about-face to focus only on AI while abandoning existing products will regret that decision.
Tim Michaud retweeted
Unfortunately I have to reshare this as I'm still looking for a position.
Had several good initial opportunities but in a lot of cases they were looking for someone less senior and not a web app SME.
If your company are hiring specifically web app testers, please DM me. 😁
I am currently looking for a remote, US-based, webapp-focused pentesting position. Though I'm not in a hurry, I understand how long interview processes can be.
If you're looking for a candidate with over 12 years experience in the industry, plus the ability to effectively communicate with customers in different roles, develop testing tools on the fly, and spread brand awareness on social media / at conferences, let's talk.
Tim Michaud retweeted
With so many high achieving people in security it’s common to feel like you never get enough work done. You should always take a step back and appreciate yourself. If you worked hard it will compound! Keep the momentum up! 💪
Tim Michaud retweeted
In this post I'll use CVE-2023-6241, a vulnerability in the Arm Mali GPU that I reported last November to gain arbitrary kernel code execution from an untrusted app on a Pixel 8 with MTE enabled. github.blog/2024-03-18-gaini…
Tim Michaud retweeted
I've audited the Android kernel in late 2023, and reported 10+ kernel bugs to Google, along with 2 exploits. Today, I'm releasing the first exploit, targeting the Mali GPU on Pixel devices, accessible from an untrusted_app context.
github.com/0x36/Pixel_GPU_Ex…
Tim Michaud retweeted
"Consumers making $45k and less" is 45% of all individuals.
In other words, MCDONALDS is now TOO EXPENSIVE for almost HALF of Americans.
McDonald's says it is seeing decreased traffic from consumers making $45k and less. Eating at home has become more affordable so the chain is losing that group to grocers, etc. $MCD
Tim Michaud retweeted
This is a five alarm fire.
SB 63 — which would criminalize bail funds in Georgia — passed the Senate today. The bill would make bailing more than 3 people out of jail per YEAR a crime for bail funds and nonprofits. It also increases the number of non-bail-eligible offenses. atlpresscollective.com/2024/…





















