Pappa Pug (@PappaPug) header
Pappa Pug (@PappaPug) profile picture

Security Researcher, Advocate & DeFi Investigator Ex @immunefi

Death City
Joined July 2023
Photo by Pappa Pug (@PappaPug)Photo by Pappa Pug (@PappaPug)Photo by Pappa Pug (@PappaPug)Photo by Pappa Pug (@PappaPug)Photo by Pappa Pug (@PappaPug)Photo by Pappa Pug (@PappaPug)Photo by Pappa Pug (@PappaPug)Photo by Pappa Pug (@PappaPug)Photo by Pappa Pug (@PappaPug)Photo by Pappa Pug (@PappaPug)
Pappa Pug retweeted
by extension: be sovereign of your private knowledge, don't put your secret recipe in someone else's database
your edge is whatever you know that the models don't know
28
36
5
833
28,127
Update: Immunefi upheld the duplicate closure. Their own fix-bypass policy says a bug that wasn't effectively fixed should be reported. That's exactly what I did. The PoC in my report still works after the attempted fix. Yet their mediation response didn't even mention the fix-bypass policy. So what's the point of having it? Why would any whitehat spend time checking whether a fix actually works if the answer is still "duplicate, no reward"? I thought Immunefi was supposed to keep this fair. Instead, they're keeping their clients happy.
Projects trying to rug whitehats seem to be the norm lately. I recently found a High severity bug. The project closed it as a duplicate of an earlier report. Which is technically true since both have the same root cause. The funny part is they already tried to fix it. And failed. My PoC demonstrates that their fix can be bypassed and the bug is still exploitable. @immunefi even has an explicit fix-bypass policy for cases like this: immunefisupport.zendesk.com/… But I guess it's easier to close the report than fix the bug and pay the whitehat.
12
4
1
111
6,463
Pappa Pug retweeted
Love this! If I had kids, I’d love to teach them this way. I’ve been exploring learning through play too! A Claude skill for making hand-drawn, interactive everyday objects. Here’s my little washing machine :) colorfigures.vercel.app
Here’s a prototype of the Kids version: “How does a pump fill a tyre?” I’m experimenting with different iterations to find the simplest way to help kids understand how things work, while still keeping enough detail to make the experience engaging, interactive, and educational
101
566
43
5,805
381,029
Pappa Pug retweeted
A two-part story (in images).
9
12
3
99
13,740
Pappa Pug retweeted
> El CEO de OpenAI admite que la IA no causará el apocalipsis de empleo que predijo > Uber no encuentra un retorno que justifique sustituir humanos por IA > Starbucks elimina la IA de su sistema por fallos constantes > Microsoft retira Claude Code a sus ingenieros por no ver mejora en costes NATURE IS HEALING
🤝 Paid partnership
JUST IN: Sam Altman says AI probably won’t trigger the “jobs apocalypse” he once predicted.
135
834
65
7,591
692,606
Pappa Pug retweeted
the top right corner of vscode is starting to resemble an iq test
157
1,276
101
26,578
573,516
Pappa Pug retweeted
really enjoyed collaborating with @AdamWhitcroft on this one!
Obsidian Web Clipper has its own icon now 🥰
18
7
3
674
46,523
Pappa Pug retweeted
The original MakerDAO code written in "daiwanese" is still one of the most underappreciated SE feasts in Web3. It's the opposite of current code slop with hundreds of unnecessary libraries and object hierarchies. Complicated code = more surface to exploit Study daiwanese. Minimize your code. Learn FV. The future is opposite to current slop practices
5
7
93
6,498
Pappa Pug retweeted
Notice how hacks increased as contests died off. You can say AI but there’s a bunch of Whitehats involved in these hacks. You onboard lots of them then you take their jobs away.
10
9
4
131
13,973
Welcome to web3 security
If you're a naturally anxious person, I recommend pursuing a high stress career path where at least you'll be compensated for anxiety you're going to have anyways.
2
1
13
932
There's a legitimate frustration buried in that take, but it rests on a misread of why BBP platforms exist. A bug bounty program is not a contact form. It's a contract. The platform is what makes that contract enforceable on both sides. Projects get filtered signal and predictable triage. Researchers get clear rules, payout SLAs, dispute resolution, and a record of work that compounds into reputation. Other responsible disclosure routes exist for the urgent stuff, SEAL 911 being the obvious one, but the premise sits in the same place. Structure exists so serious work can get through. The "just send details over email" model sounds lightweight until you sit on the receiving side. I triaged thousands of reports running MTS at Immunefi. The cost asymmetry is brutal. A researcher can write a scary-sounding critical report in 15 minutes. Invalidating it takes a senior engineer 1 to 2 hours on average. Sometimes more. Small and medium teams don't have that headcount. The big ones already outsource it. With hacks happening almost daily, projects should be taking security more seriously, and many are. But it remains a genuinely hard problem. BBPs catch what a researcher can find externally on the same surface as an attacker. They don't and can't cover everything. A lot of recent damage has been infrastructure failure rather than code bugs, and that surface is hard to fold into a BBP without monthly or quarterly pentests on top. The platform requirement does one more thing that gets missed in these threads. It self-selects for researchers willing to engage with the process. That friction is a feature, not a bug. If signing up for a platform a project explicitly named in their program rules is the dealbreaker, the report was probably not going to clear the bar anyway. Serious researchers maintain accounts on the major platforms because that's where the work is. The AI era raises the floor of all of this. Anyone can prompt a model to produce a coherent-looking critical report with fabricated impact and confident language. Valid work gets buried under that slop, which is precisely what feeds the scepticism. Structured intake, enforceable rules, and reputation on a platform are the only things that hold up at scale. At least for now. So the real question is what intake model actually gets bugs fixed at scale, not whether platforms add friction. The honest answer is managed triage on a platform, with SEAL 911 and similar channels for the cases that can't wait. Teams that skip this either drown in noise or quietly hire someone to handle it.
Replying to @v4rvl
Why provide an email for disclosures on github. I report the bug, then reach out to alert you about the severity of the bug, then you send me a link to resubmit the bug via another channel? If you don't have an account with that platform, you'll have to create one, and if maybe you can't due to some restrictions, you're blocked. Seems like too much overhead. Why not just request details? That's what most teams do. When the bug has been confirmed and fixed then we can go that route. Fixing the bug should have been the priority. What if the researcher can't send the report over the platform, they are literally saying if you can't for some reasons, then sorry we are not interested. I don't know if you get my point
2
1
8
1,495
Pappa Pug retweeted
I never realized how much of this space revolves around marketing companies and AI agents until I learned more about them. Some companies and most “AI agent” guys are just doing pure marketing nonsense, nothing more and nothing less. Meanwhile, blackhats keep showing us what real work looks like by hacking a new project almost every day. So ignore all this nonsense, focus on securing people’s money, improving your skills, and earning good rewards for yourself.
3
1
44
1,627
Pappa Pug retweeted
Future of web3sec: Arms Race - Defensive AIs vs Offensive AIs, with teams of elite humans continually improving them. Protocols that aren't monthly scanning their existing code using continually improving Defensive AIs will fall behind the curve & more likely to be exploited.
6
4
1
75
5,560
Pappa Pug retweeted
tech industry is optimizing itself to death.
JUST IN: GitLab announces job cuts to reinvest in growth for the “agentic era.”
131
5,420
64
55,856
927,919
Pappa Pug retweeted
So Mythos was, indeed, not marketing hype. Remember this is a general purpose model that just happens to be good at finding exploits because good models are good at lots of things. Expect similar from OpenAI & Google. And from open models in 8 months. hacks.mozilla.org/2026/05/be…
130
297
59
3,425
588,491
Pappa Pug retweeted
If you’re waiting until you can afford every paid AI tool before you start building, you’re going to fall behind. There is already enough free infra to ship agent prototypes: - Free Search + Fetch APIs from @Tiny_Fish - Free Cloud browsers from @browser_use - Free inference via @nvidia NIM - $5 Memory Gateway credits from @mastra_ai - Conditional ChatGPT Plus from @OpenAI You don’t need another subscription. You need to build.
🤖 Made with AI
4
30
146,842
Claude Code + Obsidian for academic research! Link to the repo in quoted post from Tom below. BTW, there's so many implementations of Obsidian + Codex/Claude Code, can you recommend one that works well for you?
Claude Code inside Obsidian for academic research github.com/hkcanan/katmer-co…
3
21
153
15,479
Pappa Pug retweeted
-refusing to ask for help -downplaying your achievements -isolating yourself instead of communicating -saying "yes" to everything -neglecting your own needs -avoiding important responsibilities -flooding your mind with negativity -always chasing perfection
self-sabotage is also:
151
13,606
555
60,871
1,328,622
The day after the CEO lays off a ton of staff and says: “Non-technical teams are now pushing code to production with AI” @coinbase has a major outage on their trading engine, and even their status page doesn’t work. 😂
Their status page is also down 😭
320
1,507
198
15,212
988,105