Joined October 2021
- Tweets40.8K
- Following2.2K
- Followers56.6K
- Likes61.5K
I'm one. It's great.
Replaced WisprFlow for me because:
- OSS
- Transcribed on-device
Only downside is no mobile client right now. May burn some tokens to change that.
Also looking to port the local speech model to other plats (e.g. my T3 nightly).
Just realized I talk to my agents the way my mom texts me.
→ Always a voice note transcription
→ Completely broken grammar and wording that isn't corrected
→ Point comes across anyways
Unsure of the lesson here. ¯\_(ツ)_/¯
The biggest reason you should all-in vibe projects is because the more you do it the more your ideas expand.
You end up in this loop
→ Do [thing]
→ Experience [thing+1]
→ "I didn't think it could do that" (expands your power scale)
→ [thing+1] becomes baseline
Before long your power growth is exp measured.
Two things:
1.) If you claimed the $100, it expires in 30d so take advantage of it while you have it.
2.) Pro move if this got you to create an account is to sign up for the free research reports alerts. Have whatever assistant/agent you have monitoring your email to grab the report and summarize bullish/bearish items for you.
NEW: Our official @OpenAI plugin is now live. Here's how to get started:
1. Head to app.blockworks.com/claim and claim your free $100 of credits
2. Connect to the Blockworks MCP
3. Let your imagination run wild
Not a ChatGPT user? Our MCP works with any other agent as well.
🚨 BREAKING: Password manager mSecure is telling current and former customers to immediately secure everything stored in it, from passwords to 2FA and passkeys.
Its maker, mSeven Software, says it "recently learned of suspicious activity" tied to the app and has called in outside cybersecurity experts.
The notice does not say what happened, how many users are affected or whether any vaults were accessed.
In December 2025, Germany's cyber agency BSI found that mSecure's maker could, in theory, access the data users store in it.
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.
From here on if you ever want someone to prove their not a NK operative, just send them to the site to get verified.
DPRKaptcha.com
Sign-in supports verifying Telegram, GitHub, X and Google accounts.
Someone wasn't happy with the site. 😅
DDoS'd w/ 8M requests.
I spend a good amount of tokens hardening the security before launch, anticipating some pushback, but they blew through my limits with this one.
Reset in a few hours and we go again 😤
This is, effectively, the stablecoin transfer thesis.
I'm here to tell you guys if you're looking for tx fees on high volumes to justify valuations that the upper bound problem is both supply AND demand.
Because the numbers required to justify even current chain vals is HYUUGE, let alone forecasting future an appreciated price (to also justify on even higher demand+supply needs).
That is what @tulipking is getting at.
Some numbers to get the point across:
Solana's ATH tx-volume week (Aug 17-23)
~2,180 TPS sustained (~92% of capacity)
~$1.05M in base + validator vote fees
If you were to sustain that for a full year, you get the below (as a company):
$58.8M in fees (a 936x multiple)
"But Bread, what about priority fees?"
Ez mafs:
$239.8M in fees (229x multiple)
You then realize that those ATH numbers are ~8% off literal max capacity for the chain (currently), so there really isn't a ton of juice to squeeze for the payment bros (agentic or enterprise) because those players WILL NOT do payment transactions if they have to pay >minimum base fee just to land. They'll wait, or opt for another chain because the cost isn't economical if it's anything but the cheapest.
"Yea Bread, but we're projecting GROWTH. Solana (and all chains) will IBRL forever."
Agreed! and not only is that actually bad for business, it shows you how far away these chains are from the capacity needed to justify the valuations.
1.) It's bad for business because IBRL directly cannibalizes a chains highest rev driver (priority fees [which routinely come in ~5x higher than base fees)
2.) Below is a chart that takes the same ATH week and projects numbers out in 2x and 10x hits. For this, you have to assume Solana will scale 10x (it will, but question is how long) and for demand to not just 10x, but 10x the ATH week for a full year.
My guess is [2] takes many years from here, but lets genie zap this into reality TODAY and you're still left with these multiples for the current $SOL MC:
Base Fee+vote: 163x
Including Priority: 21x
Now factor in FDV + inflation...
Note: This isn't a push against Solana specifically. It is easily the very top chain positioned when it comes to user demand. This commentary applies to all chains that have to justify a valuation based on fees earned. Unit economics are not there if you don't have a product.
Glass half full view here: AI future fixes this?
Everyone is mourning the death of ad revenue for companies in the agentic era because they don't view/click them on the UI like we do.
So they'll miss both these scams AND the sponsored Amazon spot for the random shit I'm buying
Now a service! Spun up an API endpoint to query test takers.
You can ask if a X/TG/Google/Github account has completed the test.
Profile pages being thrown around inspired me to finally build my own.
bread.build
I think it's much simpler: AI is eating the "dumb and/or niche question" relationship.
Reddit, Stack Overflow, etc.
Bots, and AI broadly, are eating up a huge chunk of social engagement by being really useful + having a personality.
Any platform with value derived from human data input (social media or "useful data" sites) will take a hit over the next few years as agents insert themselves into the relationship stack and we become less reliant on each other.
Short social platforms
Long technological atomization (unfortunately)
I don't see how OSS doesn't gut competition from here on out.
At this point I explicitly opt for tooling which I can see the source of because it gives me the power to shape it (either locally or formally), and I finally have the ability to with LLMs.
And usually within a day.
I already have nightly builds of @t3dotcodes and @excalidraw because I want things like voice transcription and custom font uploading, and I can just build it.
Really, really empowering.
Convos around this tweet got me to download @t3dotcodes to comp against Codex/GPT app.
I'm a convert.
Much, MUCH more intuitive, snappy and cross-machine aware.
Things I like so far:
Project management
→ Natively able to "resolve" a thread within a project so that I can keep my mental work streams clean
→ Thread status is cleanly highlighted with "Working" and "Done." Allows me to juggle my tasks quickly without having to dive in to threads to check on things.
→ Can customize projects with things like logos, default model for chats, and actions
→ Can toggle what sub-items from a given working sessions of an agent is shown (e.g. can turn off editor opens)
Cross-device
→ App support across phone/Mac/PC and easily synced natively without having to do off-app setups/installs (Tailscale).
→ It actually loads my chats fully across devices (phone <> Mac has been a PitA on ChatGPT)
Developer Experience
→ Easily juggle multiple model providers and display them cleanly per-thread/project
→ Jumping between the chat and local project files makes it a more complete developer experience/IDE than what chats offer if I want to take over (for e.g. direct copy updates)
→ Local project <> Github management flow is much more integrated, with things as simple as spawning a private repo on project creation in-app
Other
→ Opensource so I can contribute if I want something
All of the above is on the free plan, but that might not be for long if I get fully pilled.
Crazy because I always interpreted the commentary around Theo as hostile/as if he was an AI larp, but this is a really quality product.
