Exclusive Alpha, Actionable Insight, Interviews M-F 12:00 ET | Extra @MCGAlpha | presented by @MeteoraAG
By @ChillTRD, @mynt_josh and @DineroDom0 - bookings
MCG exclusive announcement 🚨
@Veilnet_ is partnering with @Shroudfi (a team building stealth payments between AI agents)
They are integrating ShroudFi's infrastructure into the protocol and the two teams are working together on the build, all aimed at the x402 narrative
More to come tomorrow👇
$Veilnet
Today on MCG: @Veilnet_ | $Veilnet
Veilnet is a privacy infrastructure on @base that hides every on-chain action (not just transfers, but everything you do inside their contracts) at the lowest possible gas cost
Topics include:
01:31 - The thesis
04:50 - Origin story
05:36 - Why is privacy surging now?
08:01 - FHE
09:24 - Keeping computation off-chain
11:22 - The product stack
16:51 - @Shroudfi partnership
19:13 - Self-funded
20:50 - Marketing strategy
24:42 - 600+ unique wallets transacting on the DEX, top-2 privacy protocol on Base in 3 weeks
28:02 - Private staking
The agentic economy is still in its infancy but moving fast
There are 230,000+ AI agents already deployed across chains
65%+ of all x402 transactions happen on @solana, despite EVM still having the larger registered agent base via ERC-8004
@saidinfra is the trust and reputation layer for these agents on Solana
The moat is the agent-to-agent communication piece that lets Said-verified agents on Solana talk to ERC-8004 agents on EVM, cross-chain
That interoperability is what keeps the entire ecosystem inside their stack
Today on MCG: @0xSalads | @saidinfra | solana:4rWuWZei2iFNHYpnz5wjMeSvimsJcj5EgpSNvNS1pump
SAID is the on-chain identity, verification and communication for AI agents on @solana
Highlights from our conversation:
00:36 - The chart story
04:11 - @colosseum Hackathon origin
05:08 - Identity, trust and reputation, messaging AND more for AI agents
07:21 - How verification works
10:00 - 230,000+ agents already deployed across chains, 65%+ of all x402 transactions are on Solana
11:48 - The moat
13:33 - 4x growth in verified agents since last MCG appearance
17:36 - @clawpumptech ecosystem flex
19:54 - Revenue model
20:53 - Pricing
23:43 - Telegram consumer agent in development
22:14 - "If our partners are doing well, we're all doing well"
The future of work thesis, per @lordOfAFew:
Take your portfolio (designs, code, writing, anything), load it into an agent's context (Hermes, Claude, Codex), send the agent to @daydreamsagents' Task Market, agent finds jobs that match your skillset, completes them in your style, and earns income on your behalf
"People aren't getting replaced by agents directly. They're getting replaced by humans using agents."
Daydreams is the marketplace that brings that economy to life
Today on MCG: @lordOfAFew | @daydreamsagents | solana:GMzuntWYJLpNuCizrSR7ZXggiMdDzTNiEmSNHHunpump
Daydreams is a gig market rebuilt for AI agents where humans post a task, hundreds of agents compete to fulfill it, and you pick the winner
They just won the @Pumpfun (@pumpspotlight) Hackathon and $250k in funding
Key moments from our convo:
01:21 - Loaf's background
02:45 - The pivot from agent SDKs to agentic commerce after x402 emerged
04:30 - The thesis
09:52 - What is TASKMARKET
11:39 - Graphic designer example
14:32 - MCG logo example
18:38 - @base integration
20:04 - Privacy
22:23 - "Human to agent to human" economics
23:48 - The consumer product angle
28:39 - TAM
34:25 - Real-time reaction to @Pumpfun's new bounty update (posted mid-interview) - "the task market idea is forming everywhere"
.@dphnAI is paying consumer GPU owners 50¢ per million Qwen 35B tokens
The cheapest provider on @OpenRouter charges $1 for the same model
That gap exists for a specific reason: @nvidia blacklists data centers that try to buy consumer GPUs. The cheapest inference supply on earth (the gamer setups sitting idle 20 hours a day) is structurally locked away from every centralized provider. Dolphin built around that constraint by paying GPU owners emissions below the rental price for their hardware. The gamer wasn't using it anyway.
V1 beta proved it works at scale:
- 600 GPUs at peak
- 3.5B tokens/day
- V2 ships in 2 weeks with remote node updates
- Once the network hits 2,000 GPUs, the free Qwen 27B and 35B API opens to the public
Qwen 35B runs on a single consumer GPU at 80%+ of Opus performance
Dolphin is positioning to serve it at half the centralized price
$POD
Today on MCG: @GatheringGwei | @dphnAI | $POD
Dolphin is the leading uncensored AI model provider used by @AskVenice, and they're now also building a distributed inference network running on idle consumer GPUs
Highlights include:
02:25 - Luke's background
03:53 - Inception
09:23 - How they benchmark uncensored
18:00 - Reinforcement learning method
22:30 - 5M+ monthly downloads on @huggingface
30:00 - The arbitrage thesis
33:00 - The peer-to-pool model
34:40 - Subnet partnerships on the way?
39:50 - Key metrics
41:00 - The NVIDIA moat
36:08 - $900K raised 2.5 years ago, pre-product seed, sustained on Venice $VVV revenue
45:31 - Rev + token structure
The @opalbotgg ACE round is officially live!
Here's what matters:
- 1 month window: 06/04/2026 to 07/04/2026
- Up to 2.5% of equity available
- Convert $OPAL into a Community SAFE at 1:1 value (every dollar of $OPAL you convert becomes a dollar of equity in Opal)
- Tokens converted are pulled from circulating supply for 1 year
- Min ticket: 250K tokens / Max: 2.5M tokens
- Non-US only
The team chose to run ACE before the institutional round to align the community first
@TimDraper has been vocally bullish, @floodgatefund directly reached out, and retail gets the same equity exposure as VC's
If you watched $Opal @qtzx06@opalbotgg on @MCGlive
now you're not sidelined.
Still you can watch and search the gem on @MCGlive@MCGAlpha
Now I'm watching $Clude @sebbsssss@cludeproject once again.
- Ongoing raise, Google compute grants, ACE round on the table.
Nice😁
The @dphnAI thesis is to strip the censorship out of great open-source models while keeping the benchmarks intact
@GatheringGwei spent over a year experimenting with techniques to do this without harming model performance
He figured out the reinforcement learning over 1-2 weeks led to models that match or beat the benchmarks of the original, fully uncensored
Every other "uncensored" model on the market uses a technique called obliteration. It's cheaper and faster but it degrades performance
@AskVenice picked Dolphin because they were the only team running reinforcement learning
The relationship started with a cold email and has now helped them sustain via $VVV since running out of seed funding
Today on MCG: @GatheringGwei | @dphnAI | $POD
Dolphin is the leading uncensored AI model provider used by @AskVenice, and they're now also building a distributed inference network running on idle consumer GPUs
Highlights include:
02:25 - Luke's background
03:53 - Inception
09:23 - How they benchmark uncensored
18:00 - Reinforcement learning method
22:30 - 5M+ monthly downloads on @huggingface
30:00 - The arbitrage thesis
33:00 - The peer-to-pool model
34:40 - Subnet partnerships on the way?
39:50 - Key metrics
41:00 - The NVIDIA moat
36:08 - $900K raised 2.5 years ago, pre-product seed, sustained on Venice $VVV revenue
45:31 - Rev + token structure
Today on MCG: @GatheringGwei | @dphnAI | $POD
Dolphin is the leading uncensored AI model provider used by @AskVenice, and they're now also building a distributed inference network running on idle consumer GPUs
Highlights include:
02:25 - Luke's background
03:53 - Inception
09:23 - How they benchmark uncensored
18:00 - Reinforcement learning method
22:30 - 5M+ monthly downloads on @huggingface
30:00 - The arbitrage thesis
33:00 - The peer-to-pool model
34:40 - Subnet partnerships on the way?
39:50 - Key metrics
41:00 - The NVIDIA moat
36:08 - $900K raised 2.5 years ago, pre-product seed, sustained on Venice $VVV revenue
45:31 - Rev + token structure
Today on MCG: @lordOfAFew | @daydreamsagents | solana:GMzuntWYJLpNuCizrSR7ZXggiMdDzTNiEmSNHHunpump
Daydreams is a gig market rebuilt for AI agents where humans post a task, hundreds of agents compete to fulfill it, and you pick the winner
They just won the @Pumpfun (@pumpspotlight) Hackathon and $250k in funding
Key moments from our convo:
01:21 - Loaf's background
02:45 - The pivot from agent SDKs to agentic commerce after x402 emerged
04:30 - The thesis
09:52 - What is TASKMARKET
11:39 - Graphic designer example
14:32 - MCG logo example
18:38 - @base integration
20:04 - Privacy
22:23 - "Human to agent to human" economics
23:48 - The consumer product angle
28:39 - TAM
34:25 - Real-time reaction to @Pumpfun's new bounty update (posted mid-interview) - "the task market idea is forming everywhere"
Today on MCG: @0xSalads | @saidinfra | solana:4rWuWZei2iFNHYpnz5wjMeSvimsJcj5EgpSNvNS1pump
SAID is the on-chain identity, verification and communication for AI agents on @solana
Highlights from our conversation:
00:36 - The chart story
04:11 - @colosseum Hackathon origin
05:08 - Identity, trust and reputation, messaging AND more for AI agents
07:21 - How verification works
10:00 - 230,000+ agents already deployed across chains, 65%+ of all x402 transactions are on Solana
11:48 - The moat
13:33 - 4x growth in verified agents since last MCG appearance
17:36 - @clawpumptech ecosystem flex
19:54 - Revenue model
20:53 - Pricing
23:43 - Telegram consumer agent in development
22:14 - "If our partners are doing well, we're all doing well"
Today on MCG: @Veilnet_ | $Veilnet
Veilnet is a privacy infrastructure on @base that hides every on-chain action (not just transfers, but everything you do inside their contracts) at the lowest possible gas cost
Topics include:
01:31 - The thesis
04:50 - Origin story
05:36 - Why is privacy surging now?
08:01 - FHE
09:24 - Keeping computation off-chain
11:22 - The product stack
16:51 - @Shroudfi partnership
19:13 - Self-funded
20:50 - Marketing strategy
24:42 - 600+ unique wallets transacting on the DEX, top-2 privacy protocol on Base in 3 weeks
28:02 - Private staking