bread.mega (@bread_) header
bread.mega (@bread_) profile picture
Joined October 2021
Photo by bread.mega (@bread_)Photo by bread.mega (@bread_)Photo by bread.mega (@bread_)Photo by bread.mega (@bread_)Photo by bread.mega (@bread_)Photo by bread.mega (@bread_)Photo by bread.mega (@bread_)Photo by bread.mega (@bread_)Photo by bread.mega (@bread_)Photo by bread.mega (@bread_)
Tired of these DPRK hacks. So I built the ultimate defense. DPRKaptcha[.]com Simple questions with existential outcomes (for them). Live now to protect the internet from North Korea.
THIS IS INSANE. In just 4 days following the biggest exploit ever, Mercado Bitcoin is hacked for $40M and an EU exchange is hacked for $106M today. Is anything safe anymore??? 🤯
165
427
128
4,214
356,484
T3codes + Omarchy is the way. Flashed both of my XPS machines this week to round out my 3 machines to enter the promised land. Spun up my own dashboard to go alongside t3 for monitoring agent activity more granularly. Feels really good operating at the OS level with agents
Man t3code has been tanking everything, my habits have improved over half my work is running on Linux. 47 threads in 2 days, everything just works.
10
25
2,294
Telegram calls just might be the single most infuriating, inconsistent way to have a conversation with someone. Truly impressive levels of bad to the point where I may reject all tg out of principle.
10
36
3,289
I pity those who fall into this glass-half-empty view of <new vs old thing>. Because it's a choice to be sad when you could so easily be happy for being able to experience that past and see good in the present. I'll never understand it.
Having one of those days where I mourn the loss of the usable internet. It's become so degraded and empty, so full of bullshit and devoid of life, so much more difficult to navigate and enjoy. It's all platforms, products, sponsored posts, scams, scams, scams.
2
14
2,290
Wut.
Contemplating my first Linux install after watching some Omarchy videos. Anyone playing with it?
15
1
41
3,709
I'm sure some of this is new OS smell, but I really like it so far
457
Another telegram widget: Turned the @mega_holdings whitepaper drop from earlier in to a game.
Any Telegram message can now become an app. ⚡ Downloads instantly (less than 4 KB)
🎨 Interactive and natively rendered
💸 Payments and viral mechanics built in 🛠️ Build your own. More info and docs for devs: t.me/BotNews/122
6
1
53
3,701
and now it's a Telegram widget. Uses Blockworks MCP to query realtime data on interactive charts.
Built this into a Chrome extension. I wasn't naturally taking research to my agents from my timeline ideas, so I brought it to X for inline research. May publish it if enough people think it'll use the $100 MCP credit
5
2
2
40
6,352
bread.mega retweeted
Today, the MegaETH Foundation became MEGA. Here's what this means for MegaETH.
Today, we commit $30 million to build new products on MegaETH, starting with AI inference. mega.holdings
67
26
13
211
32,416
bread.mega retweeted
It's Opus switching day. I made a chart that shows you which Fable model you should switch to based on how you used Opus. e.g. Opus 5.5 Max -> Fable 5.2 High is 50% cheaper for same coding output. opuswitch.bread.build
16
4
1
65
9,162
bread.mega retweeted
In early summer, after concluding our accelerator program, we redeployed our strongest engineers to a special project. They set out to find where people are exploited by the systems they depend on, and build something better. A number of prototypes emerged, and as the work progressed, one in particular kept pulling us deeper.
68
35
18
280
46,664
Contemplating my first Linux install after watching some Omarchy videos. Anyone playing with it?
28
1
39
8,525
It's Opus switching day. I made a chart that shows you which Fable model you should switch to based on how you used Opus. e.g. Opus 5.5 Max -> Fable 5.2 High is 50% cheaper for same coding output. opuswitch.bread.build
16
4
1
65
9,162
Sick - first attempt at an embedded chart and it worked.
3
7
1,593
"Fable 5.2 High scored about level with Opus 5.5 Max, for half the cost" We don't even have to worry about competitors crushing the unit economics of these labs because they are rushing to do it to themselves. Whoever hits AGI is worth infinity. The rest are very much not.
Replying to @claudeai
In early testing on FrontierCode 2.0, Fable 5.2 at High effort scored about level with Opus 5.5 at Max effort, for roughly half the cost per task. At Xhigh effort, it scored the highest of any released model tested. Medium is the default in Claude apps and Claude Code.
4
2
27
3,966
Has also become my daily setup and highly recommend. What's not mentioned: it's open source This means you can run a local fork and implement any features you want. I built our android dictation before it was live, and currently implementing an orchestrator on desktop.
I spent months of my life and thousands of dollars looking for the best AI coding setup. I've finally found it - T3 Code. So, I made a video showing my full $1,000/mo setup:
9
1
34
6,569
The future is now.
Built a physical simulation of a record player with WebGL2 and Web Audio. It's *shockingly* accurate: vinyl-gl.vercel.app. Instead of playing a song, it works by converting it into the physical shape of a groove. Then a simulated stylus rides those walls 384,000 times a second, and you hear the voltage that motion would induce in a cartridge. So the crackle, the pops, the wow/flutter, the dull inner grooves are natural byproducts of the simulation. You can also "DJ" the record. (The converted songs are in the public domain or CC-licensed. Code is linked for you to remix and re-deploy.)
3
1
22
3,022
Built this into a Chrome extension. I wasn't naturally taking research to my agents from my timeline ideas, so I brought it to X for inline research. May publish it if enough people think it'll use the $100 MCP credit
6 questions you can ask the Blockworks MCP today, answered. Then the same numbers, live in Intel. Get started with the Blockworks MCP in minutes with $100 of free credits. app.blockworks.com/claim
25
4
2
79
13,788
In general I think teams should look for simple products like this to release alongside a primitive drop. It makes the underlying more accessible and gets creative juices flowing with builders. Something the tempo team executed really well with their MPP release and has stuck with me since
1
11
913
Initial idea was to leverage decision + extraction models to pull out words from the timeline, but turns out you can be pretty precise with traditional methods. Probably implement it locally just to see how the hit rate changes on flagged words. Early iterations w/o had hits on common words like "would" (lowercase) because it's an actual token in the repo.
2
561
Starting a petition to officially change "weekends" to "vibedays" It's so easy to ship small projects I can't step away from the computer
12
35
4,560
2 of 3
Introducing Garmin Data Workbench After too many years, Garmin finally created a Personal API for you to sync your data. So I built a workbench that lets you: → Sync + archive that data → Discuss that data with agents (locally or remotely) Muse and Grokbot tested.
1
3
702
3 of 3. On to the next idea.
Built this into a Chrome extension. I wasn't naturally taking research to my agents from my timeline ideas, so I brought it to X for inline research. May publish it if enough people think it'll use the $100 MCP credit
1
336