S16 E33: Liz Steininger on Least Authority & Auditing Open Source Software
The audio keeps playing while you browse the site.
Read the transcript14,408 words, every line links to the minute it was said
Ask this episode
Liz Steininger runs a company called Least Authority, and she has a boss's job there: she took over as CEO when Zooko Wilcox, who founded the firm in 2014, left to spin Zcash out of it. The firm has finished close to 300 audits for more than 150 clients, about half of whom come back, and the client list includes MetaMask, the Ethereum Foundation, Filecoin, Polygon, Tezos, Stacks, Zcash and Keystone. It audits open source software for a living, which is odd if you believe open source means anyone can check, and its CEO tells you without hesitation that nobody can promise you a codebase is 100% secure.
We recorded this one as a livestream, with the usual mess. I stopped at the six-minute mark to plug in camera power, my partner wandered onto the screen about an hour in, and I spent part of the evening complaining that I had been without a smartphone for three days. I also opened by joking that an assistant had sent me talking points, which I then mostly ignored.
A Company Called Least Authority
The first thing I asked was whether a firm with that name has any hierarchy at all. "You're not the first person to bring that up to me," Liz said. Her answer was that "there has to be some level of authority, because that's how companies work and how they interact with the rest of the world," and that what she runs has "some structure and some leadership, but without it being authoritarian, I guess."
I then asked about Nick Szabo, whose thinking sits behind the name, and whether his writing shaped the company. Liz said she did not know the paper I was pointing at. "I don't know that paper, actually, which is sad." What she does take from the idea is a broader principle: "the idea that individuals especially have autonomy and freedom and choice over what they do is incredibly important." In her telling, the rest follows from asking what a system needs rather than what its operator would like to have, "from a perspective of what's necessary as opposed to what's wanted."
Her own route into this started young. As a teenager she sold 15 to 20 computers that used magnetic tape drives on eBay, and she learned programming through Logo. I guessed Scratch on a Raspberry Pi, which was the wrong generation and the wrong decade. Later, when she described the moment that pushed her toward privacy work, she put it simply: "That really got me to the side of privacy." She has now been at it for "about 10 years."
Why Audit Code Anyone Can Read?
I put the obvious objection to her early, with a slogan I hold dear: "don't trust, verify." If the code is open source, why does anyone need to pay a firm to read it? Liz's reply was that verification is real but unevenly distributed. Independent checking, she said, "enables autonomy for anybody who wants to double check the technology that they're interacting with," and most people who would benefit from it have neither the time nor the training to do it. "The hypothetical versus the reality of it and the practicalness of it is where there's space for businesses like us."
She is wary of the word trust here. "People don't all want to put forth the effort to do that," she said, and "the word trust is very difficult." What an audit buys, in her view, is "accountability basically at every step along the way." The auditors also see more than the public can: Least Authority works on closed-source code too, so a client's users can get an independent read of something they could never inspect themselves.
On the findings, she said her team has "found some really surprising things for sure." Clients, she added, are not hostile about it: "they want us to find the problems and tell them before somebody else finds them." She told me she misses the team-facing side when work gets abstract, and in the second half of the show she said "it is nice to get to work with teams again."
The DAO Hack and the Limits of Foresight
I raised the 2015 incentive analysis Least Authority did for Ethereum, because it came up in the research I did before the show: the firm theorized in that report about the kind of vulnerability that later played out in the DAO hack. Liz confirmed it. "To theorize that something like that could happen, definitely that came up in that report."
I asked her whether it hurts when a client ships something anyway and gets exploited. She called that "probably one of the downsides to the work that we do." Being right about a risk gives her no satisfaction, and she said so directly: "we don't get in the security business because we want to be proven right." Her explanation for why the job is thankless was that "it's hard to prove that you prevented something from happening."
Liability gets thornier. If an audited project is hacked, is the auditor on the hook? "Yeah, unfortunately, it depends," she said. The usual practice she described is to look at the delta, "looking at what actually changed" between the audited version and what was exploited. When I asked about audits for projects that did not need one, her instinct was a question for the client: "wait, what's the need?"
Nothing Is 100% Secure
This was the part of the conversation I would pull out first if I had to pick one. "This idea that something is 100% secure and that that can be guaranteed is false," Liz said. She added that "you should be suspicious if anyone tells you that you're certified secure." I mentioned that I had always found the "unhackable" claims that crypto products love to make a little suspicious, and she called it a great example, pointing out the human side of it: "if somebody tells us we can't do something, or if something's like, you know, this is perfect, and you can't break into it, somebody's going to be like, I want to figure out how."
Her limit on what an audit can promise is concrete. "Even if we manage to find all the vulnerabilities in the code, there could still be another way." And she used a password example for the human layer: "If you put it on a post it note on next to your computer, what good is the strong password?"
So what does a firm do with that? It reports what it found, helps the team fix it, and then, in Least Authority's case, publishes. About half of its work ends up published, she said: "About half of our work ends up getting published, not all of it." Client consent is required, and reports go out only after the fixes land. I said I found that disappointing for open source projects: "It's a bit disappointing given that this is all open source, but I can understand why, because they might be concerned that some people exploit some issue before the team gets to fix it first?" Liz said the delay is for that reason, and she was also open about the limits of publishing a score at all: "but the security score is just not enough."
Sharded Ciphertext and ZKAPs
PrivateStorage is the product I wanted to understand properly. It runs on Tahoe-LAFS, the old distributed storage system, and adds ZKAPs so that the people running the service cannot connect a payment to the storage it bought. Liz's description of what ends up on the servers is short: "basically sharded ciphertext." Because of that, "you don't have to trust the PrivateStorage service at all."
The ZKAPs are the privacy layer on the payment side, designed "to ensure that the payment information that we get from people doesn't get connected to their usage of the storage service." In her version, the company can see that someone bought storage and that storage is being used, "we just can't connect it to who bought the storage time."
I asked whether this generalizes. She thinks it does: "any service where the service provider doesn't have a fundamental need to know your identity." She also said Tahoe-LAFS does not require a company at all: "you and a bunch of friends could all install Tahoe-LAFS inside of your own P2P network and, you know, store on each other's machines, etc. And you would still get the client-side encryption and sharding."
There is also HRO Cloud, which gives storage to human rights organizations. Liz said that for some of those orgs the work has been "just unpaid, for those orgs." I asked how the numbers work for a firm that does this and also has to run a business. "We also have to pay rent and stuff like that," she said, and she described the tradeoffs between funding models bluntly: "all models have their tradeoffs for sure."
Winden and the Wormhole
Winden is a file transfer tool built on the Magic Wormhole protocol. When I asked what people actually use it for, Liz gave the most mundane answer: "one really popular use case is just sending something between your laptop and your phone, for example." The sender gets a short code, which she designed to be shared aloud, "so that you could say it over the phone or if you're in person together."
There is a catch she did not hide: "both devices have to be online at the same time to receive it." That is the price of not parking your file on someone else's server.
Maintenance is the other theme she kept returning to. People, she said, "want to see some sort of shiny new thing," while "just maintaining open source stuff that people like, it's not that exciting." I asked her about a book for the ZK curious. She told me there is one she wishes existed, "a textbook for regular people to read about how ZK works," because the subject "seems like magic, and it seems like kind of out there, but it's real and it works."
Collecting Less
I asked how privacy squares with how companies make money. Liz thinks the conflict is overstated: "The interests are not always at odds with each other." Her main tactic is minimizing. "If we don't collect certain data, then we don't need to be prompted to share that data," she said, and Least Authority tries "to minimize what we collect or just not collect anything unless we have to."
She asks one question of every feature that wants user information: "do we really need this information?" On regulation she was more open than I expected from a security firm. "Having a regulation to tell tech companies to not collect as much personal data seems helpful at this point," she said, adding that it has "helped to keep in check some of the big tech companies that are just collecting data, you know, like, yeah, way too much data about us as individuals."
Zero-knowledge proofs, in her view, have a business case too: "This idea that you can prove that you have something, that you know something without actually sharing the data itself is so valuable from a business perspective." Whether companies are doing enough is a different matter. "Whether or not they're legitimate or they're going as far as they need to," she said about the efforts she sees, and of the progress so far, "but not as much as I would like to see."
Auditing Vibe Coding, and What We Still Argue About
I asked about the market for auditing AI-generated code, since half my guests this season have been vibe coding something. She has not seen the rush yet. "We haven't seen like a spike yet," she said. Her take on the demand is that it depends on whether anyone cares: "do you want to go hire somebody to audit something that you just did for fun?" What she does want is "the quality check on what are we getting when we vibe code."
What her team likes most is the work it was built for. "We really enjoy new applications with advanced cryptography," she said, and she called them "a kind of good challenge for us because we've done so many of those over the years." She did not paint her own industry as a stable one: "Firms kind of come and go," she said. "There's good years and bad years for everyone." I also corrected myself on air about Sergio Lerner's firm, which is Coinspect, not Coinfirm. I mixed up the names and clarified that Sergio Lerner's company is Coinspect, not Coinfirm; Sergio's earlier visit was S15 E7 on Rootstock, Satoshi's Bitcoins, Drivechains and more.
She is optimistic about where this is going. "We're still just so early in all of this," she said, and she keeps herself calm by looking back at an earlier internet: "I try to stay grounded by thinking back to, like, the early days of the internet. And what was it, Gopher? And then, like, how that, you know, changed over the years." Her conclusion for digital value transfer is that "privacy has to be front and center." My own memory of that era is AOL, and she remembered it too: "I remember using AOL."
Where we still do not agree is on publication. Liz publishes about half of what Least Authority audits, with client consent and after the fixes, and she thinks that is how responsible disclosure should work. I think that when the code is open source and the audit is paid for, the public deserves the report whenever the client would rather keep it quiet, and I told her I find the gap disappointing. She sees a timing problem, where a public report before a fix hands attackers a map. I see a trust problem, where unpublished audits ask users to take the firm's word, which is the thing I thought we were trying to stop doing.
Listen to Liz Steininger on Spotify, Apple Podcasts, YouTube, and More
Time stamps:
- 00:00:00– Intro and Sponsor Mentions
- 00:01:17– Guest Introduction: Liz Steininger
- 00:01:57– Irony of "Least Authority" Having a CEO
- 00:03:04– Least Authority Philosophy and Nick Szabo’s Influence
- 00:05:19– Liz’s Tech Background
- 00:09:36– Role of Auditing Firms in Open-Source
- 00:11:45– Surprising Audit Findings
- 00:12:16– Notable Clients and Audits
- 00:14:35– Predicting the Ethereum DAO Hack
- 00:17:43– When to Conduct Audits
- 00:19:51– Auditor Liability and Security Guarantees
- 00:22:25– Social Engineering and Security
- 00:29:55– Least Authority’s Products: Private Storage, ZKAPs, Winden
- 00:36:45– ZKAPs Applications Beyond Storage
- 00:43:53– Winden’s Features and Use Cases
- 00:46:21– Destiny: Mobile Version of Winden
- 00:50:00– HRO Cloud for Human Rights Organizations
- 00:51:00– Moon Math Manual
- 00:57:38– Privacy, Law Enforcement, and Zero-Knowledge Proofs
- 01:01:59– Privacy Market and Big Tech Adoption
- 01:05:00– Auditing Non-Crypto Projects
- 01:07:17– Vibe Coding and Auditing Demand
- 01:11:07– Zcash Audit and Bitcoin Codebase
- 01:13:05– Publishing Audit Reports
- 01:18:01– Competing Auditing Firms
- 01:21:12– Future of Privacy and Zcash Integration
- 01:24:00– Liz’s First Computer and Early Tech Memories
- 01:28:37– Closing and Winden Recommendation
Verify this episode
This is the SHA-256 fingerprint of the MP3 as published here, timestamped in the Bitcoin blockchain with OpenTimestamps. If one byte of the file changes, the fingerprint changes.
- SHA-256
3d0179d4f8bd18bc5784a98c352b0ef45150c6cee4a7471182ebb7df7240e557- Proof
- 2025-07-09_s16_e33__liz_steininger_on_least_authority_open_source_software_auditing.mp3.ots · all episode hashes (proof)
- Status
- Anchored in Bitcoin block 974,312
- Timestamped
- After release. The episode came out on 9 Jul 2025, but this proof was made on 3 Nov 2026 and anchored in block 974,312, so it proves this exact file existed by 3 Nov 2026, not on the release date.
How to check it yourself
- Download the MP3 and the proof into the same folder.
- Run
sha256sum 2025-07-09_s16_e33__liz_steininger_on_least_authority_open_source_software_auditing.mp3(on a Mac:shasum -a 256 2025-07-09_s16_e33__liz_steininger_on_least_authority_open_source_software_auditing.mp3). The result must match the SHA-256 above. - Install the OpenTimestamps client (
pip install opentimestamps-client) and runots verify 2025-07-09_s16_e33__liz_steininger_on_least_authority_open_source_software_auditing.mp3.ots. It hashes the MP3 again and checks the proof against your own Bitcoin node, then prints the block and the date the file existed by. Without a node,ots info 2025-07-09_s16_e33__liz_steininger_on_least_authority_open_source_software_auditing.mp3.otsshows the block height to look up in any block explorer.
Related
S17 E45: Talking Tari with Naveen Jain, Fox & Marguerite de Courcelle
There are projects in this space that are famous for what they shipped. Tari used to be famous for what it has…
S17 E29: Grafton Clark on Vibe Coding, Privacy & Bitcoin Culture
Some podcast episodes are about a product, a protocol, or a company. This one is about a feeling: the feeling that the…
S17 E24: Zach Herbert on Passport Prime, KeyOS & the End of Air-Gapped
Some interviews you plan for weeks. This one I planned around a box. Two weeks after Zach Herbert handed me a Passport…
Comments
Comments run on my own server. No trackers, no tracking cookies.