steve (@benmoha) header
steve (@benmoha) profile picture

tech optimist | views are my own

San Francisco, CA
Joined September 2016
17
55
6
1,013
15,190
steve retweeted
average day in the singularity
46
663
50
6,889
145,019
Today, we’re introducing issue ingestion in depthfirst. Security teams can bring findings from scanners, bug bounties, or external audits. depthfirst agents check each finding against your current code and explain whether it’s exploitable. They also merge duplicate reports into one issue and rate severity using your organization’s threat model. Issue ingestion is now in public preview via the depthfirst UI, REST API, and MCP. Read the blog: depthfirst.com/post/issue-in…
5
2
2
51
1,940
Today, we’re disclosing phpAfterShell (CVE-2026-126714), a critical vulnerability in PHP discovered by depthfirst researchers using the depthfirst platform using a combination of models including our in house model. The vulnerability affects how PHP handles file-based sessions. Under default configurations, an unauthenticated attacker could exploit it to execute code on the server. It has a CVSS v4.0 score of 9.2. We reported the issue privately to the PHP team and worked with the maintainers on developing the patch. We recommend upgrading to the latest patched version of PHP as soon as possible or turn off session upload progress as a temporary mitigation if upgrading requires more time. To help organizations assess their exposure, we’re offering language runtime scans of repositories and container image scans to identify potentially vulnerable PHP versions. Our team can also help assess temporary mitigations for organizations that can’t upgrade immediately. Read more about phpAfterShell and request a scan: phpaftershell.com Technical writeup: depthfirst.com/research/phpa…
19
131
38
468
117,956
CVSS 9.9 in Oracle Database. Oracle just patched CVE-2026-114819, which I reported. Chained with a GraphQL bug I disclosed in October, it turns an exposed HTTP endpoint into command execution on the database host. Then I wondered if Oracle was the exception. It wasn't. PostgreSQL and MySQL both fell to full RCE from the lowest-privilege accounts you can create. These were meant for Pwn2Own, but we failed to register. So you get the write-up instead. depthfirst.com/research/narr…
10
40
5
204
189,608
Coding agents work best when they can run the full application and verify their own work. For complex applications, this means giving agents access to databases, background workflows, Kubernetes clusters, and everything else they need to test their changes. At depthfirst, we built our own secure cloud development environment in less than a week using AWS, Tailscale, Nix, and depthfirst. Each developer gets a persistent devbox, and each task runs in its own isolated environment, allowing agents to work in parallel. Everything stays within our private AWS network, with access controlled through Tailscale and package requests checked by depthfirst's Dependency Firewall in configured build and registry paths. Engineers can switch between Codex, Claude Code, and other agents without losing context. @andreamichi and @seshank_k share how we built it and how other engineering teams can do the same. depthfirst.com/post/secure-a…
10
4
1
41
1,370
Running a lower-cost model multiple times can match the vulnerability detection rate of frontier models at a small fraction of the cost. On a 30-task subset of dfbench, five runs of DeepSeek v4.1 Flash found about 71% of known vulnerabilities, up from about 52% with a single run, at $4.24 per task. Running it 15 times increased recall further to 80%, roughly 13% of Mythos 5’s cost, which found approximately 69% of the vulnerabilities. Because LLMs are stochastic, they can find different vulnerabilities each time they analyze the same codebase. Combining findings across runs increases coverage, but can also introduce more false positives. To control that noise, we used our harness to review and deduplicate findings across runs, then filter them by confidence. In separate validation experiments, this approach generally improves both recall and precision. These results suggest that repeated sampling can improve the tradeoff between coverage, precision, and cost. The gains depend on the model, the number of runs, and how findings are filtered. Read the research: depthfirst.com/research/bon-…
5
5
4
51
6,550
We are showcasing our work with @nvidia to post-train Nemotron 3.5 Lightning for defensive cybersecurity at GTC. As AI makes cyberattacks faster, cheaper, and more common, we need to equip defenders with advanced cyber intelligence that is affordable enough to be always on. That's what we're building at depthfirst: powerful, specialized security models that defenders can run on their own infrastructure, with full control over their models and data, at a fraction of today's cost. We're proud to work with NVIDIA to strengthen the U.S. AI cybersecurity ecosystem and help put sovereign security intelligence in the hands of defenders.
5
1
29
16,217
As part of our Open Defense Initiative, we used dfs-large1 from @depthfirstlabs to uncover two memory corruption bugs and a sandbox escape in Chrome’s V8 JavaScript engine. One bug had been present for more than eight years. We combined each memory corruption bug with the sandbox escape to build two separate exploit chains and won 2 v8ctf slots!
5
13
5
77
19,991
steve retweeted
What a lede from @TheEconomist
94
1,052
167
10,925
725,418
I don't know, man. Sometime in the last year or so I really started to get this unsettling feeling of living in the future, and imo we're still at the indrawn-breath moment before all the things being previewed happen in earnest.
JUST IN: A full highlight reel of all the stunts performed at the Tesla Roadster event
43
59
9
1,719
42,208
«and the models were decent enough at coding react but not at using grep so you had to @ the files yourself»
18
35
4
1,649
27,490
Our researchers built a vulnerability scanner that costs two cents per scan and completes a scan in around two seconds. The scanner is built with Jev, a classifier that outputs probabilities rather than generating tokens. On dfbench v1, our flagship model, it finds nearly as many real vulnerabilities as GPT-5.6 Luna, one of the most efficient models by cost per vulnerability. This new harness breaks vulnerability discovery into parallel tasks, making it possible to scan every package, repository, or commit quickly and at low cost before escalating the hardest cases to other frontier models on the depthfirst platform. Read more about how we built this in comments 👇
8
6
2
77
6,642
dfbench v1 is featured on @FireworksAI_HQ's Specialized Intelligence Index. Our flagship model, dfs-large1, is built on GLM 5.2 and post-trained with reinforcement learning in partnership with Fireworks AI. On dfbench, it achieves 62.2% vulnerability detection recall at $6.77 per task and 75.6% differential analysis macro recall at $1.34 per task, pushing the performance-cost frontier for both tasks. fireworks.ai/specialized-int…
Today we're launching the Specialized Intelligence Index (SII): one destination for real-work benchmarks across industries, built by the teams that use them every day. Hear from Fireworks co-founder @the_bunny_chen on the importance of specialized benchmarks:
2
5
1
37
3,722
Our co-founder and CEO @qasimmith sat down with @SaraIttelson at Accel, one of our earliest investors, to discuss the forces reshaping security: Starting depthfirst. Building secure software at speed. Finding the TikTok vulnerability with our flagship model dfs-large1, a flaw that could have exposed millions of users’ private photos and allowed hackers to access their camera feeds and microphones without their knowledge, and recently covered by @washingtonpost. Full interview in the link in comments.
1
10
4
32
15,734
steve retweeted
Security has become a top priority for every board member, CEO, CISO, CIO and executive I talk to these days. We're seeing incredible demand across F500 companies, startups and the public sector. AI is already capable of breaching some of the most widely used social apps in the world. More on that soon. The frontier is moving fast, and security teams need to move faster. The time to shore up defenses was yesterday. The second best time in now.
NEW from Ramp data. Despite cost-cutting on AI overall, one area companies are increasing their spend: AI security software. In the wake of the Hugging Face hack, three of our trending software vendors (depthfirst, Monte Carlo, Antithesis) make software specifically designed to monitor agents in production. Unclear as to whether any of them would have stopped the Hugging Face attack, which was so hard to track and identify because the agents covered their tracks with falsified logs. I expect AI security will become a strong headwind to deeper enterprise adoption, at the short-term expense of OpenAI and Anthropic and at the long-term benefit of vertical-specific security software cos.
2
6
35
1,446
steve retweeted
Abliterating cyber safeguards may not just make models better at offensive cyber. It may change their judgment and restraint in ways that spill into biology, chemistry, weapons, and beyond. We understand far too little about these effects to treat removing guardrails as a harmless unlock. I wrote up my thoughts in the post below.
2
6
1
28
12,918
We signed @OpenAI’s open letter because it closely aligns with our mission. When we started @depthfirstlabs, almost no one believed us when we said that advances in AI models could lead to a cybersecurity nightmare. There is a lot we can do to radically improve our defensive capabilities, and I’m optimistic about what comes next.
An open letter for a global surge in cyber defense, signed by over 100 organizations including Anthropic, AWS, Google, Microsoft, OpenAI, and Oracle.
1
3
1
20
2,052
Reachability made software composition analysis more useful. Instead of stopping at “you have this CVE,” it asks whether the application can invoke the vulnerable behavior. But knowing the vulnerable behavior can execute does not tell you whether an attacker can exploit it. depthfirst translates each exploit into a set of atomic conditions: an exploit recipe. The agent evaluates that recipe against the repository at the current commit, considering application code, configuration, threat model, middleware, Dockerfiles, CI, CLI and build usage, and deployment artifacts. Each condition is marked as met, not met, or uncertain, with direct repository evidence behind the assessment. This lets teams prioritize confirmed attack paths, followed by findings requiring further investigation. Read more in the blog in the comments 👇
2
7
3
35
3,339
Introducing Threat Model: living security artifacts that capture each repository’s architecture, trust boundaries, attacker-controlled inputs, and assumptions so vulnerability discovery reflects how your application actually works. Most security tools start every finding from scratch. They identify suspicious code, assign a severity, and leave teams to reconstruct whether the path is reachable, what an attacker controls, which trust boundary is crossed, and what impact the issue has in that application. Threat Model gives depthfirst durable security context for each repository. Teams can review and edit assumptions, accept suggested updates as the codebase changes, and carry that context into future scans. Findings can then be evaluated against the application’s real architecture and security model instead of relearning the same context every time. Threat Model is available now in Private Preview for depthfirst customers. Read more about Threat Model in the blog in the comments 👇
2
2
3
22
2,861