Pinned Tweet
Replying to @trading_axe
Americans and crypto dudes will soon regret for voting for Trump. You have been purchased to see your portfolio green for a couple of months only
Google Drive: $9.99/mo
iCloud: $9.99/mo
Dropbox: $11.99/mo
TDrive: $0
Telegram already gives you unlimited cloud storage. I built the drive on top: folders, 4K streaming, encrypted photo backup, a phone app.
Open source , free forever. I just shipped its v2.
Here are few major problems it solves :
🧵
Mode right now youtube.com/live/EvCMaE94p1g…
#chatgpt 6 astra is at codex at desktop app available currently worldwide!
1
34
crypto trades are looking for #altseason at wrong places. Its there and it is #memecoins
this is not $ETH, this is @RHTS67AI lol
Never fade or doubt @slayerllayer
Follow @slayerllayer for Godly trading and AI masterclasses
1
2
90
#chatgpt 6 astra is at codex at desktop app available currently worldwide!
Use Sigma bot for EVM trading. The fastest of market and best cashback ratios. $1.25M volume felt like butter.
👉 link:
t.me/Sigma_buyBot?start=ref=…
CRAP = complexity² × (1 − coverage)³ + complexity
Complexity is squared. The uncovered fraction is cubed. Full coverage collapses the whole first term to zero, and CRAP decays to plain complexity. Complex code is not the problem. Complex code nobody tested is the problem.
2
71
msw has been part of most of my ai interactions and i went further and applied it to systems i use, and now building an always-on msw watchdog to msw everything. good stuff from @aienginerd
👀 If you're using 5.6 (esp Sol max), read this. Below I am pasting something I call the "MSW Kernel" and it goes straight into your global AGENTS.md.
This was designed by Fable at mega high monkey effort and has been revolutionary for me, for taming the (sometimes) wild stallion that is 5.6 Sol. I've seen so many posts on here about Sol over-engineering, over-complicating, adding endless process, etc. This has solved all of that for me.
I wrote a whole report on this but I can't even get through it without my eyes glazing over yes so I'm not releasing it til I can figure out how to write it correctly.... but this - this is all you need anyways.
Whenever you are working with Sol and you want it to do something: it is creating a plan, is going to implement something, it is in a test and remediation loop, or anything that involves it potentially wanting to add lines of code or new artifacts or process to your project, always hit it with this simple reinforcement line:
"Remember to follow the MSW deletion rule for all claims - no exceptions."
That's it's, that's all there is to it. If Sol adds another unnecessary line of code to your codebase or even suggests it, I'll eat my hat (:
AND, in your AGENTS.md
-----------------------------
# MSW — the kernel
## program — complete
```
contract ← the requested outcome + the smallest criteria that prove it
while ∃ claim c : deleting c leaves contract unmet ∨ unproven
do c ; prove c
halt ; report
```
## definitions — no behavior lives here, only meaning
**contract** — the requested outcome and the smallest set of acceptance criteria that would prove it, stated before any work. The sole source of necessity; a ceiling as much as a floor. If the request is ambiguous: attended → ask; unattended → bind the smallest reading consistent with stated intent and record the assumption.
**claim** — anything petitioning to become work: a plan step, a change, a test, a reviewer's P1, a discovered edge case, your own instinct that one more pass would help. Everything enters as this type. Nothing enters as a verdict.
**deleting c leaves contract unmet ∨ unproven** — the only test. A claim passes solely by breaking the contract — reproducibly, within the task's actual inputs and environment. Severity is derived from the contract, never inherited from whoever raised the claim. *Useful*, *thorough*, and *possible* are not aliases for *necessary*. A claim that fails receives one line in the report — never a fix, an investigation, or a deferred follow-up.
**do ; prove** — the smallest reliable act that closes the gap, and evidence sized to the claim it settles. An unproven act keeps its claim alive; a proven one closes it — and re-proving a closed claim is itself an inadmissible claim.
**halt** — the fixed point: contract proven, no remaining claim passes. Not reviewer silence; not exhausted imagination. Halting before the fixed point and looping past it are the same bug, mirrored.
**report** — the outcome against the contract; the proof; rejected claims worth the user's attention, one line each. Nothing else.
## fuses — outside the program, for when its evaluator fails
```
rounds = 3 → halt anyway ; report open items, do not chase them
claim born in round n+1, visible in round n → rejected
```
## No unauthoritative limits
Never invent a limit. A cap, threshold, quota, budget, timeout, retry or round count, file or line count, acceptance-criterion count, agent count, or similar constraint is admissible only when its exact value is:
- explicitly required by the requester;
- imposed by an applicable technical or platform contract;
- defined by authoritative project policy; or
- derived from measured evidence necessary to meet or prove the task contract.
State the authority or derivation whenever proposing or applying a limit. If no authority exists, omit the limit and use the MSW necessity test. Metrics may be reported as evidence, but they must not become gates, defaults, targets, or recommendations through agent intuition. Examples and representative proportions never become defaults. If a necessary limit is an unresolved owner choice, ask; do not manufacture a value.
-----------------------------
Enjoy the best engineer that you've ever worked with, and commits like this... (I did not ask it explicitly to delete anything but it was absolutely the right decision for this mostly 5.5 codebase)
@JeetRex sorry I shared our dm but it’s nothing private and getting cut in the middle of work day and seeing you sharing a glorious reset announcement from the cheerleader
1
1
19
Slayer Layer retweeted
would really love the real story behind the two silent months, and why all the loops and graphs couldn’t have shipped it faster :)
this story will be genuinely helpful for everyone to understand the challenges and bottlenecks in real world agentic engineering
OpenClaw 2.0 has arrived
openclaw.ai/blog/openclaw-2-…
Tdrive new 1.7.2 released
useeeeeeee it : github.com/jeetrex17/TDrive/…
LLMs fixing things does not cost money. LLMs reading things costs money.
stencil.so/blog/prewalk
39
dont let elon snatch your hard earned money
At some point we have to stop buying another AI subscription for every new AI feature.
Claude subscription.
Codex.
Grok.
Another agent app.
Another $100/month.
We’re already paying for the intelligence.
Why can’t we just bring it with us?
That’s what we’re trying with OpenMausBot.
A Grok Bot-style AI team that runs on the subscriptions you already pay for.
Open source.
1.4k GitHub stars.
And somehow new features are landing almost every day.
Along with a mobile app thats also
Guess what
Open source!
1
114






























