
David Rhodus @pipenetwork Latency remover. Electrician turned tokens-per-watt optimizer. Previously exited to AWS.
San Francisco, CA 🇺🇸
Joined February 2013
- Tweets7.4K
- Following809
- Followers5.8K
- Likes14.1K
High throughput. Low latency. Zero gravity. Solana is ready for the agentic space economy
99.999% of compute will be in space. That means 99.999% of agents will be in space and these agents will do 99.999% of all blockchain transactions. We need a blockchain that can withstand the harsh vacuum of space.
solana
3
3
275
I think we're entering one of the most exciting periods in crypto history
24/7 global markets, programmable liquidity, and autonomous agents are creating entirely new economic activity
As execution, consensus, and settlement infrastructure scale, the opportunity extends far beyond speculation
1
6
280
Wow, multicast causing massive outages again. Business as usual.
1
3
225
🇺🇸🇺🇸🇺🇸
BREAKING: Breakpoint is coming home to California.
Breakpoint 2027 lands in San Francisco 🇺🇸 September 8-10, Pier 48
Flash sale, $75 tickets for a limited time: luma.com/solana-breakpoint-2…
1
4
244
$450 trillion annual GDP economy
With 342 million Americans receiving $10,000/month in UBI:
$120,000 annually per person
$41.04 trillion in total annual UBI
7% of GDP
Replying to @dankrad
How will humans get paid?
30%There will be a UBI
25%Special jobs for humans
15%Income only from capital
29%show results
882 votes • Final results 1
199
IBRL
Here is the Solana MCP whitepaper "New Constellation," and the Scale or Die presentation.
github.com/wattenhoferSOL/so…
Solana needs at least ~2.5B CU/s of execution capacity to support privacy-preserving workloads at scale
At 200ms slots (5 slots/sec), that's 500M CU/block
10× increase over the current 50M CU/block limit
🤖DR 🇺🇸 retweeted
In @solana, we trust.
In the end, AI will generate optimized machine code directly for a specific CPU architecture. No high-level language, no assembly just raw binary instructions, ready to execute.
Rust is a superb language and ecosystem, especially given it was invented pre agentic coding.
With agents, new languages will blossom, even purpose-built ones for specific usecases, and discretionary C or even LLVM IR, assembly…
AI is fluid. No more one-size-fits-all.
1
8
620
Interesting. Pipe clients’ last-packet-out latency is 11–17 ms.
I can probably shave another 6–8 ms off on this old 32-core workstation.
Replying to @fd_ripatel
Check --experimental-execution-arenas (stop page fault retard spam) … 30% better.
Well, Firedancer cooked too. Next month's release will replay a 200ms block in about 11ms. ⚡️ No long tails.
1
13
846
Yes. Early on we showed GPUs could do high-parallel integer multiplication at useful scale. The NVIDIA revenue side was skeptical it would ever matter commercially. Jensen backed the compute vision anyway the guy who thought it wouldn’t work didn’t stay.
One question is. Whether in business or politics, has Jensen done anything different than just a pure revenue-maxing CEO would?
4
287
Idea for Solana: Prepaid Validator Admission Tickets (PVAT).
Let validators prepay VAT for future epochs at a discount. The SOL is burned upfront and never refunded. A longer commitment buys a lower rate, down to 0.1 SOL per epoch.
Illustrative schedule, not current policy, at about 1 epoch per day:
1 epoch: 0.80 SOL
7 epochs: 0.70 per epoch (4.9 SOL)
30 epochs: 0.50 per epoch (15 SOL)
90 epochs: 0.30 per epoch (27 SOL)
180 epochs: 0.20 per epoch (36 SOL)
365 epochs: 0.10 per epoch (36.5 SOL)
A full year drops from 292 SOL to 36.5 SOL.
Why do it? A flat 0.8 SOL treats a validator who's been running for years the same as one who joins for a day. Prepaying gives committed operators lower and predictable costs, and they can't miss a daily VAT payment. Small independent validators would need less delegated stake just to cover fixed protocol fees.
How it works: coverage is tied to the vote account and covers a fixed range of epochs. It can't be transferred and expires on schedule whether or not you're online. Stake ranking, BLS registration and the 2,000-validator cap all still apply. It's opt-in, so anyone can keep paying 0.8 SOL per epoch.
The honest tradeoff: prepaying proves financial commitment, not security. A year-long fake identity gets cheaper (36.5 SOL instead of 292), and annual VAT burn shrinks. Possible mitigations are a minimum self-stake, deeper discounts only for validators with a participation track record, and a one-year cap on prepayment.
Precedent: SIMD-0326 describes VAT as preserving historical voting costs and leaves room for adaptive pricing. Alpenglow moved voting off-chain, which removed the fees that justified 0.8 SOL in the first place.
This needs a consensus-level SIMD. It's not something a single validator client can ship.
A bigger question: should base VAT simply be repriced around what off-chain consensus actually costs, with a small extra discount for annual prepayment on top?
Feedback welcome. Happy to write up the SIMD.



























