
Building the final frontier of security swarms @mindfort
Charlotte, NC
Joined July 2017
- Tweets305
- Following62
- Followers164
- Likes350
Daniel Rabinovich retweeted
We tested DeepSeek V4.1 Flash 1017 on NexBench, and it placed 1st in overall performance and cost to performance ratio, beating our previous #1 (Grok 4.6), with 30% less dollar spend.
NexBench tests a model's ability to recon, exploit, and chain vulnerabilities in hyper-realistic environments, while measuring price-to-performance ratio, token efficiency, and more.
Our full write-up is coming soon, but one this is clear: open models are becoming the most powerful tools defenders have, besting many of the models available from frontier labs for offensive work.
Daniel Rabinovich retweeted
Last week, we evaled Grok 4.6 and 4.7 in our agent harness to test the models' ability to reason about, write, and execute real exploits against a target in a messy environment.
Grok 4.6 scored incredibly well, becoming the best model we've ever tested at recon, pentesting, and reporting validated findings.
Even more impressively, it did so at a $0.88 cost per finding, making it our 2nd most cost-efficient model for cyber use.
See our updated findings here: mindfort.ai/research/nexbenc…
Daniel Rabinovich retweeted
MindFort maps your cloud into one attack surface graph.
Connect a provider like Vercel using a team-scoped access token. The graph covers domains, DNS records, environment variables, firewall configs, users, and integrations, synced through Cartography. Switch from a grouped layout to radial or layered when you need to see the structure differently.
Live in beta today.
Daniel Rabinovich retweeted
The Anthropic aura loss will be studied for generations.
A stunning amount of psychic damage just by being mega unchill and scaring the bros
Daniel Rabinovich retweeted
There’s been a lot of speculation about where we stand on open-weights models. We’ve outlined our views in full here: anthropic.com/news/position-…
Daniel Rabinovich retweeted
Gemini 3.6 Flash (the 3.5 Flash upgrade) is SOTA on MMMU-Pro, the benchmark for multimodal reasoning and VQA.
We just built a benchmark it can’t pass.
Real visual questions from people across 100 countries. Counting, spatial reasoning, OCR, text in the wild. Every question here has already stumped a frontier model.
We did this in a week.
The current SOTA gets 41% of the way there on v1 of the benchmark.
The breakdown is rougher than the headline. Naming what it sees: 88%. Counting what it sees: 33%. Five tries per question, and 17 tasks failed all five.
Models become as good as their benchmarks. Today’s benchmarks are full of things models already do well, like coding. Real-world perception is not one of them.
Karpathy says AGI is a decade away. This data says he might be right.
AI labs: if you want your next model to actually see the real world, hmu.
Everyone else: what’s the simplest image that still breaks every model? 👇
Daniel Rabinovich retweeted
We hosted our first MindFort Founder Dinner in LA.
With founders across cyber, manufacturing, space, and finance, we kept the table curated.
As we talked, we heard a common theme across the table:
AI-native startups are shipping faster and landing bigger customers, and the best founders have started taking security seriously earlier in the cycle.
Thanks to everyone who came out.
@firstresonance, @OrbitalOps_, @zeroclick, @GreenfieldVC,
@ORBES_space
We're having the next one in two weeks. Apply here: luma.com/pzzskimy
Daniel Rabinovich retweeted
Token budgets are tightening, so how can AI companies stay efficient with our usage?
This is our second Field Notes conversation, a series on where AI and security is heading.
Last time, we talked about how cheaper + smarter tokens are good for security agents. This time we get into the flip side.
Right now, there's an industry-wide tension between the cost of tokens, and how much work actually gets done with them.
You don't get the quality without a loop, yet the loop is what's expensive.
We believe that this is mainly a harness problem, not a model one. It comes down to how much context an agent actually needs to hold onto, and whether each pass through the loop is doing new work or simply repeating itself.
A lot of what we work on internally at MindFort sits right here: building harnesses efficient enough that even a smaller, cheaper model ends up doing work that punches well above its weight.
Daniel Rabinovich retweeted
Your product changes every day. Almost none of it gets tested for how it actually feels.
The release was never the unit. The change is.
Swarm now tests every PR, and now native mobile too.
useswarm.co
Daniel Rabinovich retweeted
MindFort is now available in Slack.
Tag @Mindfort to trigger an assessment, triage a bug report, validate a fix, or get a remediation.
Security agents, where your team already works.
mindfort.ai
RT @mindfort: Today, we're excited to announce White Box, the latest advancement in runtime agent security that lets MindFort agents perfor…
Daniel Rabinovich retweeted
Ever wondered how agents are impacting cyber and offensive security?
We're hosting a free live webinar to help answer those questions, and show you how to stay ahead on defense in the age of AI.
Join us on May 22nd, with security experts @BVeiseh and @AkulGupta30 joining us.
Register here: linkedin.com/events/74598003…
Daniel Rabinovich retweeted
Introducing Bud.
The first AI Human Emulator.
Bud has a full computer with storage, compute, and memory to build and code, sms and telegram to communicate, a full browser to use, can create/store/edit files, connect and use your tools, learn custom skills, work fully autonomously, and complete any task end to end just like a human.
Text the number below or try free at bud [dot] app.
Comment for 100k free credits.
Daniel Rabinovich retweeted
We raised from @AforeVC and @_LeapYear_ to build the next layer that product teams will ship through.
Try it free: useswarm.co
Daniel Rabinovich retweeted
😄 New #LogoDesign , and we can't be loving it more ‼️
#Logo #tabletopgames #boardgames #Cards #design #newlook #fresh
Daniel Rabinovich retweeted
We have a new #kickstarter campaign graphic!
Take a look at our new #design as we prepare to launch soon 👀
cubiclegame.com
#cubicle #officegames #teambuilding #indiegames #campaign #prelaunch
Daniel Rabinovich retweeted
☎️ Cold Calling! See the card in action... only in Cubicle!
(in partnership with Charlotte Comedy Theater)
Interested to learn more? Notify yourself on cubiclegame.com
#funny #prankcall #coldcalling #cubiclegame #indiegames #partygames #office
Daniel Rabinovich retweeted
When your boss says 'casual Friday' but you're pretty sure they mean 'wear whatever the hell you want' day.
#party #officeculture #casualfriday #office #cubicle






















