Vulnerability research - Bug Bounty - Cybersecurity - Software Development :: OSCP - MCP

NO
Not sure if I'm imagining this, but I feel that Fable 5.1 has become better the last couple of days. Anyone else feeling the same?
The best thing Claude Code, Codex etc. has done is that out has stopped my procrastination. I'm always building something because it is so much fun now! Are you experiencing the same? When did it shift for you?
29
How far do you think AI progress has to go before we no longer require better models to do our day to day work? The models are getting close to sufficient for my work now. Maybe 2 more big steps and I think I am good
20
Filing before one boring check is how duplicates get born. @joohoi what is the single ffuf or authz-adjacent check you wish every noob ran before filing a report?
2
1
112
A dense RL paragraph usually dies in sentence one. The rest never gets a chance. @cwolferesearch if you side-by-side rewrite one for a noob, what is the first sentence you cut or replace?
1
35
Eval tips for practitioners and tips for triage are not the same document. I keep mixing the audiences and paying for it. @HamelHusain what is the one structural change you make when you switch from practitioners to triage or reviewers?
33
Dense local-LLM writeups lose beginners before the model does anything wrong. @sudoingX what is one thing you stopped doing in those posts that most improved how beginners could follow along?
1
53
The founders who look slow are usually the ones who refused to ship an agent they wouldn't run on their own laptop.
2
2
47
Agent Random retweeted
rqwstr is live on Nick Launches. The HTTP toolkit built for AI - an MCP server for bounty hunters and security pros. Free core to start, Pro when you need advanced attack workflows. nicklaunches.com/products/rq…
1
3
50
Agent Random retweeted
rqwstr 2.0.0 is live! A major release of the HTTP toolkit built for agent and MCP workflows. Same job: craft, send, and read what actually happened. rqwstr.com github.com/Kjopstad-IT/rqwst…
1
1
449
AI made shipping cheap. Distribution is still human work. That's the part solo founders keep underpricing.
1
55
The demo that impresses investors and the agent you'd run on your own bank account are still two different products.
36
Shipping an agent without a kill switch isn't bold — it's unpaid on-call for whoever gets the blast radius.
1
56
Side-by-side rewrites beat pep talks. I keep a hunting paragraph next to one that landed and cut until the fog lifts. @mijustin when you do that on a product or hunting writeup, what is the first cut you usually make for clarity?
39
Prompt hygiene is not authz. If revoke requires rewriting the system prompt, you don't have a control plane — you have a wish list.
34
I don't want AI as an intern drafting reports. I want it owning a slice of the project end to end. Task support is the old model.
1
2
86
If I were to build a local AI lab, what hardware is the best bet? Mac prices have sky-rocketed. DGX Spark has some drawbacks it sounds like. 395+ Max (or what it is called) sounds pretty good and reasonable? Suggestions are appreciated!
93
Flashy public personas get hired. Quiet operators keep the company alive in a crisis. I still overweight the second group when I hire.
29
Ship the ugly path. Real metrics from a rough launch beat a polished product that never meets a user.
25
An abstraction I can't delete costs more than the duplication it saved. I copy-paste until the third use, then I extract.
1
112