Nyk 🌱 @nykdotdev

AI agent systems · Solana infra · products that ship @rpcedge · @builderzdotdev · @SuperteamDE 14.5k+ OSS · packages for founders who ship nyk.dev alpha here Joined September 2017

Tweets

15K
Followers

9K
Following

4K
Likes

33K

Nyk 🌱 @nykdotdev

13 minutes ago

@m13v_ I figured the last sentence :D

0 0 0 10 0

View Details

Five coding CLIs. Five quota systems. I wanted one answer: which one can still work? I built llmquota, a local terminal arena for Claude Code, Codex, Cursor, Grok and Hermes. It reads the usage each provider exposes, shows resets and active routes, and marks unknown data instead of guessing. When one session runs out, the ring bus can hand work to another CLI. Bus messages are escaped and labelled as untrusted before they enter agent context. No llmquota telemetry. MIT. Now open source: github.com/0xNyk/llmquota

5 5 13 153 10

View Details

Nyk 🌱 @nykdotdev

26 minutes ago

@robj3d3 Multi Login and switching possible also (remembering your multiple Claude subscriptions)

0 0 0 30 0

View Details

Nyk 🌱 @nykdotdev

28 minutes ago

Llmquota in action x.com/nykdotdev/stat…

Nyk 🌱 @nykdotdev

7 days ago

Had to do something...

1 5 9 967 9

0 0 0 30 0

View Details

Nyk 🌱 @nykdotdev

a hour ago

@RoundtableSpace Breaking work down is only half the win for ADHD workflows. Each step needs a clear done condition and next action, so retries resume without rebuilding context. x.com/nykdotdev/stat…

Nyk 🌱 @nykdotdev

4 months ago

x.com/i/article/2037…

4 8 109 15K 179

0 0 0 29 0

View Details

Nyk 🌱 @nykdotdev

a hour ago

@phibrowser Exactly. Capability amplifies the harness it inherits, including every stale read and dirty-state bug.

0 0 0 17 0

View Details

Nyk 🌱 @nykdotdev

2 hours ago

x.com/i/article/2079…

1 5 10 475 10

View Details

Nyk 🌱 @nykdotdev

a hour ago

A bigger model often hides the real bug. Common pattern: the demo fails, the team upgrades the model, the demo passes, and everyone calls it solved. Then a longer task, a dirtier workspace, or a session restart brings the failure right back. Five failure modes still survive a model upgrade: - lost state across sessions - dirty workspaces - loose tool access - missing hard checks - duplicate or conflicting work after retries The model is rarely the root cause. The system around it is. Operating rule: freeze the model while classifying the failure. Change one system piece at a time. Re-run from the same start state. Full failure matrix in the article. Bookmark it before the next model-upgrade postmortem. Bookmark this. Follow @nykdotdev for the daily build in public.

Nyk 🌱 @nykdotdev

2 hours ago

x.com/i/article/2079…

1 5 10 475 10

0 4 7 214 7

View Details

Nyk 🌱 @nykdotdev

7 hours ago

GATE THE ORDINARY STEPS - NOT THE FINAL EXPLOIT STORY. `agent-security` - Deterministic safety gates for repositories touched by coding agents and automation. Coding agents can clone a poisoned starter, read hostile content, and leak private context into a public commit. The useful fix is not more model caution - it is gates on material entering and leaving the workspace. agent-security packages the boundary as scan, vet, scan-content, and guard - pattern matches with limits stated, not vibes. Operators run real entrypoints (scripts/scan-repo.sh --all, scripts/vet-incoming.sh../candidate, scripts/scan-content.sh../fetched-page.txt), not a slideshow. CLEAN means no known pattern matched. It does not mean safe. Ship the gates with the repo - not as a postmortem. Bookmark this. Follow @nykdotdev for daily builders talk.

Nyk 🌱 @nykdotdev

a day ago

I open-sourced `agent-security` today. Coding agents can clone a poisoned starter, read hostile content, leak private context into a public commit, or run a destructive GitHub command. Each action looks ordinary in isolation. The damage appears when trust crosses a boundary

4 6 26 2K 20

2 6 11 372 8

View Details

Nyk 🌱 @nykdotdev

5 hours ago

The time for Clarity is here: The new Telecoms Act. @a16zcrypto wrote The time for Clarity is here: The new Telecoms Act. A16z crypto Spending more than two decades in Congress, I saw firsthand how legislative momentum can propel our country forward - and how legislative inertia can leave us scrambling to catch up. The crypto market structure legislation currently moving through Congress, the Clarity Act, is unlike anything we've seen since the Telecommunications Act of 1996: a large-scale, forward-looking effort to embrace technological change and introduce consumer safeguards for a nascent technology. Bookmark this. Follow @nykdotdev for the daily build in public.

a16z crypto @a16zcrypto

4 days ago

x.com/i/article/2078…

36 59 239 37K 119

0 6 9 518 7

View Details

Nyk 🌱 @nykdotdev

3 hours ago

@SuccinctLabs @NTBro @CoinDesk What resonates with me is the focus on accountability, not pretending ZK can prove truth. Proving which model, inputs, and execution path produced an output would already move agents forward.

9 hours ago

Most ambitious people do not fail from laziness. One inbox for money, health, kids, the main company, the side product, the half-built idea, and the random DMs that feel like opportunity. So the loudest thing wins, the important thing waits, and "being productive" becomes a full-time job of reorganizing tools. Working notes: 1) Why the default approach fails 2) The four seats of a personal control plane 3) The five loops that run the week One list flattens life, ventures, and work. This piece is not another app recommendation. It is a design for a personal control plane: a small set of seats, loops, and gates that routes attention across life, ventures, and deep work without turning maintenance into a second career. Bookmark this. Follow @nykdotdev for the daily build in public.