exploring agents. prev: stablecoins at @nativemarkets/@stablecoin/@stripe, consumer/social crypto, @opensea, seaport co-author | he/him | views are your mom's
Los Angeles, CA
Joined May 2021
- Tweets5.7K
- Following3.4K
- Followers5.5K
- Likes11.7K
4 is the unlucky number in east asia
tetraphobes vindicated once again
chinese century
Announcing 3 Critical vulns for hundreds of millions of IoT, Firewalls, OT, and Military devices including: Palo Alto Networks, Comcast, D-Link, HP, Canon, Honeywell, GE, Oracle, Siemens, Rockwell, Digi, Lockheed Martin, Boeing...
The vulns are with GoAhead Web Server which is a backend library used by these devices. There's a patch available from GoAhead but that doesn't mean there's a patch available for the devices yet -- that's up to the manufacturer.
- TLS traffic over DHE ciphers have a 1-byte secret, "04", meaning anyone can decrypt connections to these devices and steal login credentials and session tokens if the client negotiated with DHE ciphers like the latest TLS_DHE_RSA_WITH_CHACHA20_POLY1305_SHA256, which is extremely likely.
- LibreSSL mTLS has been completely bypassed, meaning anyone can connect over mTLS. If you have one of these devices listening on the internet and are expecting mTLS to protect it from attackers, take it offline or patch asap.
- OpenSSL Certificate Revocation Lists (CRLs) are not enforced, meaning if a user device is stolen or compromised, and you revoke that mTLS cert, nothing actually happens. There are a lot of other attacks this can be used for.
These are critical because these types of devices run our world. Water systems, electricity, military, firewalls, critical infrastructure.
foxio.io/blog/choose-your-ba…
why is grokbot voice transcription so bad? uniquely bad, doesn’t seem to use on-device transcription 🤔
got an M5 Ultra 256GB Mac Studio that i need to decide if i want to keep
hit me with benchmarks you wanna see
there's no reason to run glm-5.3-flash when deepseek-4.1-flash-1017 is both smaller and smarter, right?
nothing has wasted more of my agents’ (and my) time more than unnecessary skills
Anthropic is now recommending that companies REMOVE unnecessary subagents from their AI systems.
This is a pretty significant shift in agent architecture, with their team finding that a single agent equipped with skills can frequently outperform complex multi-agent systems, and one customer seeing roughly 2x faster execution.
The emerging rules of thumb:
→ Skills: Teach the agent HOW to do something, including procedures, expertise, and instructions.
→ Tools / MCP: Give the agent the ability to DO things through APIs, actions, and live data.
→ Subagents: Use for parallel work, isolated context, or tasks a single agent struggles with.
The interesting part is progressive disclosure. Instead of spawning specialized agents or stuffing everything into a giant prompt, one agent discovers and loads skills and tools as needed.
Their advice is to fold existing subagents into the main loop, turn their instructions into skills, expose their actions as tools, and run evals to see if performance holds up.
As models have gotten smarter, improving agent architecture is becoming less about what you can add and more about what you can delete.
claude.com/resources/article…
TIL “leech” and “leech” are not only spelled differently but have different etymologies
huh. guess i just figured that leach was a metaphor for, like, sucking minerals or chemicals out of something
restricted my coordinator agent's tools to just ask, task, and wait to force it to delegate all work, since it steadfastly refused to listen to instructions to delegate everything
it just started using the cua-driver mcp to do stuff instead 😐
been meaning to do this
my goal is to have all my tailnet devices (+ ideally remote provider apps like grok or chatgpt) sync all sessions to a single devices, which then indexes them and makes them searchable via qmd or zvec
can both act as memory and training data then
We should stop keeping all of our agent conversations local. They should be stored in a shared place.
This lets you do some powerful things:
- Schedule self-improvement agents to review runs and pull out failure modes worth fixing
- Replay past conversations against different models for benchmarking
I met a team of devs zipping up Claude and Codex sessions from their computers and shipping them to a shared S3 bucket. That's... a start 😄 But really, all agent sessions should happen on a single source (like a cloud runner) so that conversation traces are stored together by default.
It's why Warp Factories uses cloud sandboxes and makes sessions available via API. We also added an inbox view to see which conversations need your attention, which is quite convenient:
idk why but i like that english has a lot of phonemes. a lot of its vowels don't exist in other languages. that's cool
its spelling could use an overhaul though
dipping back into claude w a team account.
the unforgivable thing that claude is especially heinously bad at
claude: do you approve me to do x y and z?
me: tell me about y first
claude: *does not respond because it’s doing x and z*
my og apple tv 4k plays weird with my home theatre and sometimes i need to power cycle either it or the amp to work
i've been waiting 3 years for apple to release a new apple tv so i can see if it goes away lol
hell yeah
Write-up coming and you all get it here first. A personal project I made over the weekend -- turn any piece of analog recording equipment into a VST plugin using Hermes/etc.
You don't even need passive equipment in front of you to get it working.
github.com/somewheresystems/…
been innovating behind the scenes on a lifestyle hack i am calling the “shower white monster.” can’t wait to share it with the world
i think assistants should have a vm and most other things should not
