lil retard (@comic) header
lil retard (@comic) profile picture

don’t listen to me I’m retarded

Joined August 2012
I've changed how I work over the last few weeks. I now have one lead agent for each area of focus, running all the time, and a chief of staff agent that oversees them. Each lead manages its own sub-agents that write code, test it, chase reviews, and coordinate with team members. Everything I need to decide or review comes into one prioritized queue, a simple local web app. Each product review includes screenshots and a video demo of the feature. I clear the queue in batches, mostly by voice and keyboard shortcuts. The leads also bring back ideas I hadn't considered. Working from one queue, instead of switching between dozens of chats, has cut a lot of mental overhead. The system keeps learning my preferences and distilling my judgment into a "Brian brain," so its recommendations and first drafts get better before they reach me. The goal is fewer corrections per change. Every change still goes through the same human review and security checks. It is eliminating tasks, not human judgment. What an exciting time to push the frontier of productivity, especially as we build the economy for agents (agentic finance). We should all keep pushing in this direction.
753
410
331
8,938
1,256,330
lil retard retweeted
Founders always ask me what product market fit looks like. Here you go:
we just reached >2000 active paying users for boat.dev's cloud VMs! 2x in less than 2 weeks!
9
4
1
293
22,187
lil retard retweeted
I’m sorry but…Pattinson walking around like a suave badass, banging hot chicks, beating the piss out of fools, dropping one liners, drinking martini’s, letting his natural British accent fly? With that music behind him? And Villeneuve’s direction?? Who TF doesn’t want this?
94
1,019
69
16,857
222,974
lil retard retweeted
Congratulations, Bitcoiners. Six years in, the biggest Bitcoin buyer in the world is sitting at his own cost basis, around $75,500. Saylor promised 30% a year. Gold would have doubled his money. The Bitcoin president just got rejected by the voters. Got gold? youtu.be/eVYj3d_Nni4?utm_sou…
268
72
18
895
126,066
CANCEL your weekend plans. The 2027 meta just dropped. You NEED to: • Build a computer-use agent that parses the DOM, diffs screenshots and maps clicks to coordinates so it can drive legacy software with zero APIs • Ship generative UI where your LLM returns live React components instead of markdown blobs and watch your product feel 10 years newer • Wire a System-One router that resolves 95% of requests in milliseconds for fractions of a cent, escalating only the hard 5% to reasoning models • Kill your `while True` loop and rebuild it as a Temporal state machine that resumes from the exact node after a crash or a 3-day approval • Write a raw MCP server in pure JSON-RPC so any external agent can call your product with scoped credentials, no SDK required • Add grammar-constrained decoding so your agent emits 100% valid JSON on the first try, every try, forever • Build a trajectory grader in CI that blocks the merge the second a tool-call sequence deviates from the golden path • Ship an inter-agent sanitization proxy that executes prompt injections before they cross a single agent boundary • Wire a PII unmasking proxy that swaps sensitive data for UUIDs before inference and restores it after with full audit logs • Add a cost kill-switch that halts any loop projected to exceed $0.50 and degrades frontier → mid-tier → local → cache • Implement KV-cache prefix sharing at the gateway and cut Time-To-First-Token by 80% on repeated system prompts • Set up speculative decoding: local 3B drafts, frontier 70B verifies, acceptance rate logged on every single request • Shadow-route 5% of production traffic to your new model, diff trajectories and cost silently, auto-post the report to Slack • Automate the DPO flywheel: thumbs-down → preference pairs → nightly LoRA → eval-verified promotion, zero human touch • Build a voice loop with VAD and sub-1s interruption handling, streaming tokens and tool calls simultaneously • Ship an API-to-browser fallback that spawns a headless browser the moment a REST endpoint 404s, finishing the job anyway • Add a WebGPU shim so a 2B model runs in the browser when the cloud dies and your users never see a spinner • Fuzz your own agent swarm with malformed tool outputs and context overflow until it self-heals without a single pager alert • Publish one benchmark: your system vs the naive baseline, on cost, latency and task success, fully reproducible The wrapper era is over. The systems era pays. You have way too much to do. Bookmark & Repost.
51
103
14
1,545
164,341
lil retard retweeted
Companies that offer integrated solutions for local inference will boom in the next few years. I'm not talking about another GPU in a case. Hardware + OS + an inference stack that makes open models easy to run. A plug-and-play system. Ease of use is still underrated.
123
97
24
1,697
168,450
lil retard retweeted
Subagents are the biggest cause of AI failures In most harnesses: 1. Main agent writes a summary of what it knows 2. Subagent starts from that summary only, and makes assumptions 3. Subagent works and sends back a new summary of what it did Same flaw as compaction: all summaries are lossy. This is how your assistant makes emails you never wanted, or purchases what you told it just to get the price of So I tried something new: checkpoint agents The initial agent has complete context of what needs to be done. And that input is already cached Fork the session where it is. Send the fork ahead to do the task and it will do exactly what the main agent would have done; their context is identical What is even better is when the fork finishes, the cache of the main agent is kept warm by the subagent (they share the same prefix) I measured it on a 137k-token session with Opus 5.5: → Main agent does it on its own: 67 s, $0.38, context +8-17k → Checkpoint: 117 s, $0.68, context +2k → Classic subagent: 123 s, $0.57, context +4-5k But here is what I found: avoiding large input is even more important than keeping cache. And most subagent tasks were done better either in series with retrieval, or split up to be parallel over fresh sessions from the start I do not think there is a place for subagent delegation at all. There should either be a super-agent that orchestrates and does no work, or a long running agent that can retrieve from previous sessions, not a worker that completes tasks partially and delegates the rest to other workers Don't design harnesses based on our assumptions of human behavior
189
119
60
1,863
208,625
lil retard retweeted
71.6 million new $MSTR shares were generated out of thin air (just like the fiat fractional reserve debt based monetary system) since $STRC dropped below par in May. That is over $7B of capital (mostly dilution and lower BPS); and some people think $100M stock buyback is going to make a dent in the common stock price destruction that @Strategy has inflicted on its shareholders?
Replying to @TheRealDHR
Just the announcement of $100M buyback of $MSTR funded through $STRC would be much appreciated by common shareholders. The common shareholders paid for $2B of STRC buybacks, now its STRC time to Pump our Bags!" @BritishHodl do you concur? @saylor @phongle @CJ_Bitcoin @rohanhirani
8
1
1
38
7,454
lil retard retweeted
good prompt "find all of the implicit state machines in my app"
149
355
234
7,993
1,720,654
lil retard retweeted
Over the past six years, $MSTR has been the biggest buyer of Bitcoin in the world. Yet the current price of Bitcoin is lower than its average cost. That means all @saylor has to show for six years of stacking is a loss. MSTR would have been much better off buying gold or the S&P.
239
77
20
1,152
135,697
What just happened? The Nasdaq 100 just posted its largest daily drop since the 2020 Pandemic, with the S&P 500 erasing -$3 trillion today. Just 24 hours ago, the Dow surged +1,600 points after the Democrats "swept" the Midterms. So, what changed? Let us explain. (a thread)
276
1,184
189
8,148
2,025,774
lil retard retweeted
I don’t think the political apparatus of either party truly understands how much the US economy depends on AI succeeding. If the US AI growth story slows or stops, we will have a global macro crisis. The debt burden cannot be fixed without major reforms, which are politically unpalatable. The market is sending a warning today…
34
27
7
249
19,536
lil retard retweeted
There were a plethora of posts on the timeline yesterday celebrating $STRC returning back to par. I still believe Strategy is no longer a Bitcoin company. It's a dividend company. And $MSTR is nothing more than a piggy bank in the ponzi that Saylor uses to fund dividends. Notice how BTC Yield has been down only since June, despite both the reserves and the price of BTC going higher. This was THE metric they used to define success and shareholder value for MSTR... e.g. when they could sell more shares of MSTR to buy more Bitcoin and it be accretive rather than dilutive... until it stopped working. Then they made up a new metric with new numbers for what's actually accretive and dilutive and never mentioned it again. They currently have ~$4.9B of cash in their reserves. How did they get it? - primarily by selling billions of dollars of MSTR atm while it was still dilutive to shareholders And remember, they've also REPURCHASED $2B of STRC while it was at a discount. I actually think that's smart! But, remember the preferred stock is what got them into this mess in the first place... And now that it's back to par, they're simply going to add back to the debt by issuing more STRC to buy more Bitcoin which adds back to the dividends, obligations, ad infintum. Not to mention, there's a big chance $MSTR itself is about to see billions of dollars in sell pressure if its removed from MSCI indices. While I don't see there being a situation anywhere in the near term they'll be forced to sell their entire Bitcoin stack if the price goes low enough, I do think it's obvious they've shown they care much more about protecting the dividend than they do preserving accrual of BTC for the average shareholder... which is what the initial purpose of the company was supposed to be. So I still encourage anyone who buys MSTR because they're bullish on Bitcoin to reconsider. If you're bullish on Bitcoin, don't buy a stock that can be printed infinitely to pay someone else. Just buy more Bitcoin.
32
4
1
71
25,148
BREAKING: US technology stocks collapse at the open, erasing -$1.3 trillion in market cap in 30 minutes, with dozens of names down 10% or more. Tech Stocks Today: 1. SanDisk, $SNDK: -15% 2. Western Digital, $WDC: -13% 3. Nebius, $NBIS: -13% 4. AMD, $AMD: -12% 5. Seagate, $STX: -12% 6. Qualcomm, $QCOM: -10% 7. Micron, $MU: -10% 8. Tesla, $TSLA: -10% 9. Dell, $DELL: -10% 10. Arm, $ARM: -10% 11. Marvell Technology, $MRVL: -10% 12. CrowdStrike, $CRWD: -10% 13. Intel, $INTC: -9% 14. CoreWeave, $CRWV: -8% 15. Palantir, $PLTR: -8% 16. SpaceX, $SPCX: -7% 17. TSMC, $TSM: -7% 18. Nvidia, $NVDA: -5% 19. Broadcom, $AVGO: -5% 20. Oracle, $ORCL: -5% The Nasdaq 100 is now down -4% on the day. We expect a highly eventful day ahead. Turn on our post notifications @KobeissiLetter for real time analysis as this develops.
283
690
285
6,184
930,760
open your competitor's homepage. right click. view page source. ctrl+f "GTM-" that's their google tag manager ID. copy it now paste this into your browser: googletagmanager. com/gtm.js?id= and add their ID on the end a wall of code opens. it's public. it has to be, every visitor's browser downloads it. it's the file that runs their entire tracking setup now ctrl+f "AW-" every google ads conversion they track. if tere are 6 of them, they're optimising for 6 different things. most teams can't even tell you which one matters now ctrl+f "event" their conversion names, typed by their own marketers. "demo_booked." "trial_started." "sql_qualified." you're reading their funnel in the order they built it now ctrl+f "value" sometimes a number sits right next to the conversion. 250. 800. 1500. that's what a demo is worth to them. they typed their unit economics into a public file now combine all four: -the GTM ID gets you into the container -AW- tells you how many things google is chasing for them -the event names give you their funnel -the values give you what each step is worth their tracking, their funnel and their numbers. from one file they can't hide. every head of growth reading this just opened their own container now take their demo value. if they pay $800 for a demo and you know yours is worth $2,000, you can outbid them on every keyword you both want and still win (9 years inside google. tracking is the first thing we open on every audit → link in bio)
118
318
21
5,682
492,371
I think we are already cooked. The results of Opus 5.5 is insane.
Everything I’ve posted lately was made with a tool I’ve been quietly building. Today it’s public: aligncanvas.com beta is live. Turn your images into 3D motion in a few clicks. Nothing to install, and your images never leave your device. It all runs in your browser.
51
96
17
2,797
174,546
lil retard retweeted
I asked my agent to build a collection of our favorite designers, called Deimos. Each article was written through multiple rounds of discussion and refinement between different models. deimos.evilrabbit.com/
63
65
14
1,533
72,250
lil retard retweeted
If you design with AI, this is all you need 1. studio.motionsites.ai — cheaper than lovable 2. motionsites.ai — premium website prompts: 800+ 3d websites, motion sections. Paste into Claude, ship. 3. land-book.com — landing page inspiration done right. 4. recent.design— cool gallery for web design inspo 5. landdding.com — top notch landings 6. fontinlogo.com — fonts behind brands 7. godly.website — award websites 8. 60fps.design — mobile UI animations save these for your next project
54
268
13
2,235
130,114
Fun fact, at a $1.5TRN Anthropic IPO price, FTX's $500M from the 2022 Series B would be worth $39B. Ouch.
15
7
2
170
27,676
lil retard retweeted
Looks like Saylor is gonna be a problem again * sep '27 he owes $1B if he can't get MSTR above $183 * mar '28 he owes $2B if he can't get MSTR above $433 * jun '28 he owes $1.5B if he can't get MSTR above $672 * then there's another $2B+ in potential payments out to jun '29 ^ Assuming he cannot recover the share price, MSTR will be out of cash by March of '28. But even worse, they're only 5 months away from having less than the 12 months cash in reserve they promised to maintain earlier this year. *expect them to charitably interpret their USD reserve policy moving forward > So why assume he can't get the share price where it needs to be? Honestly maybe he gets bailed out by Bitcoin on his upcoming September converts. But keep in mind Microstrategy has had negative Bitcoin yield this year. BTC per share is down 3.7% this year. Common holders are getting brutally diluted and market stress makes this worse (BTC yield was -11% in Q3 alone). With this class of management, I do not see how Saylor convinces anybody Microstrategy should trade at a premium to Bitcoin - which means to get MSTR up ~40% to $183, you're going to need to see Bitcoin well above ~$107k. Getting this pig back to $433 then $672 for the two following converts seems essentially impossible to me unless Bitcoin goes on a sun run. > So can't he raise more cash and kick the can down the road again? Unfortunately the powers that be have decided they've had enough of Saylor and he's getting kicked out of the indices. Expect the problem to only compound as the conditions get worse
Horrible 10-Q. Flags short-term liquidity needs for cash repayment of convertible debt, acknowledges likely exclusion from MSCI indices, flags this could trigger rating revision & adversely affect the ability to raise capital. No wonder the CFO quit last week.
107
58
22
622
100,293