Allie Howe
@vtahowe
Followers
1K
Following
18K
Media
385
Statuses
4K
CEO of Growth Cyber | SOC2, NIST AI RMF, AI Security Gap Assessments | Helping AI Startups Build Trustworthy AI | Host of the @insecureagents Podcast
Joined January 2015
👋🏼 Hi, I’m Allie! Realized I haven’t introduced myself on this platform. After leading security at a startup from founding to acquisition I’m now a vCISO for startups. Reach out to me for help with - SOC2 - creating a custom security roadmap for your business - AI security
7
1
80
NYC @insecureagents live recording with @samuel_colvin @leonardtang_ @ProulxKerem @ianlivingstone after @aiDotEngineer CODE Summit cc @vtahowe @swyx @ai_pensar @haizelabs @pydantic @KeycardLabs
1
7
25
Big day ahead at @aiDotEngineer Code Excited to discuss MCP, coding agents, and how AI Engineers are approaching AI Security Great to see some fellow AI Security friends @joshdevonai @bcrussett
2
4
35
and to give credit where credit is due I stole this graphic from
arxiv.org
The rapid development of large language models (LLMs) has led to the widespread deployment of LLM agents across diverse industries, including customer service, content generation, data analysis,...
0
0
3
Reminder that MCP isn't the only agent protocol out there Likely the most popular and controversial since it's one of the few general purpose protocols Excited to touch on this during the debate
2
1
8
We’re flattered 😊 Going for an A++ on this MCP debate we’ll record and publish as an episode
Reminder to cybersecurity podcasters that no one is forcing you to do an episode on AI. So far no one has gotten above a C on an AI episode except @insecureagents, and AI is their whole thing.
1
0
6
Huge opportunity 👀
i'm hiring 5 members of technical staff from the 5 frontier labs in the next 5 weeks if you or a talented loved one is from: - openai - anthropic - google deepmind - xai - meta …i’d love to connect reach out: fivefrontiers@haizelabs.com
0
0
2
The MCP Debate is going down this Thursday at 2:30pm @aiDotEngineer CODE @dexhorthy challenges MCP, @ianlivingstone defends Is MCP > a bad protocol > rotting the context window > really the best way to do tool calls Find out Thursday!
3
8
43
Chinese Models! Preview of the Great MCP Debate - Agents Hour, Nov 17th
0
1
2
Often times security is just a scoping issue >Do you really need every tool an MCP server comes with? >Do you even need an MCP server? >Do you need that amount of non determinism that comes with an agent or can you use an LLM workflow More capabilities, more problems
3
0
2
Prepping for the great MCP debate In Anthropic's 2024 blog post announcing MCP they said, "[MCP's] aim is to help frontier models produce better, more relevant responses" first potential debate question, sooo how is that going :)
3
0
8
Giant doge at the cowork today. I’m told paw on the foot means he likes you 🥰
1
0
9
After doing compliance for AI startups for a year I'm starting to wonder if point in time audits hold any real value LLMs change as you interact with them, MCP servers change tools without your approval Real time technical controls and continuous compliance are better
1
0
13
As of right now we’ve got a solid 1v1 debate lined up I’ve got some ideas for how I could expand this and include more guests in the debate but I’d need more MCP defenders If you’re interested in defending MCP reply here or DM me
@AAAzzam @dylan522p @jefrankle @vtahowe @ianlivingstone @jlowin @vtahowe now has more guests than she know what to do with!
4
1
13
Excited to host this AIE Code side event. Hope you can join us! Luma link in this tweet or find it on the @aiDotEngineer CODE website
Join us for a LIVE discussion next Thursday after @aiDotEngineer CODE Hear from @ianlivingstone @leonardtang_ @ProulxKerem @samuel_colvin on > the current state of AI security > building trustworthy AI > securing coding agents Luma link below, RSVP now to save your spot
1
0
10
It’s always easier to be jaded or skeptical Harder to say this doesn’t work quite right but I’m interested in fixing it
1
0
7
ok one of the things that i've always wanted an AIE is coming to pass, after the Great @dylan522p v @jefrankle debate of 2024: the Great MCP debate! @vtahowe and @ianlivingstone are taking on all challengers - if you are a knowledgeable MCP skeptic, come do a live debate next
Some say MCP is >obsolete >a security nightmare >creates unnecessary abstraction Others say MCP provides >standardization >a security boundary at the network layer >federated data integration Which is it? Time to host a debate, live at @aidotengineer CODE next week!
17
8
81
For anyone that would like to see my face more closely I’m doing two live podcasts at AIE Code next week 1. MCP debate at the conference 2. Panel convo with security leaders as an evening side event on Thursday Both of these will be very high signal, clear view of my face 😄
Podcast review: @insecureagents by @vtahowe. AI agent security stuff. Solid B. Content is strong. Allie matches the technical depth of the guests better than anyone. But they talk too much about the guests’ companies. Bonus points for video, minus points for mic blocking face.
2
0
14
Some say MCP is >obsolete >a security nightmare >creates unnecessary abstraction Others say MCP provides >standardization >a security boundary at the network layer >federated data integration Which is it? Time to host a debate, live at @aidotengineer CODE next week!
4
5
41
I do think sandboxes have the potential to make some MCP use cases obsolete I can’t quote two tweets at once but I liked @masonwilliams post from yesterday saying “the primary benefit of the MCP protocol is to give agents a way to interact with 3rd party tools in a way where the
Sandbox infrastructure is going to be play a big role. - Everyone is slowly coming to terms that models are great at writing code. - MCP tool calling or general tool calling in a loop quickly becomes unreliable and hardly works for tasks involving large amounts of data
1
0
7