FhtAbd retweeted
"The Simple Mathematics of Large Language Models" is a concise free paper if you want to understand the mathematics behind modern LLMs and systems such as ChatGPT and Claude.
I have shared several resources in the past that explain the mathematics behind these systems in detail, but they are often quite long. This paper is only about 20 pages, so I think it is a useful resource if you want a compact introduction without starting from a much larger textbook.
It covers token representations, context, attention as learned weighted averaging, projections and bilinear forms, multi-head attention, normalization, positional information, softmax, maximum likelihood, cross-entropy, and gradient descent. Together, these topics give a concise mathematical picture of how a language model works.
Save it to your bookmarks. It can be useful as a quick reference.
crc.business-school.ed.ac.ukβ¦
FhtAbd retweeted
Someone went on-chain and verified everything about $TIG. (See below)
Before you read it, here are the two questions everyone asks first:
Q: Was there a VC round?
No.
No named investor.
No fund that received a discounted allocation.
Ever. Publicly announced.
Q: How was the token launched?
Not on a launchpad.
Not on a CEX with a waitlist.
Not with a presale batch.
Zero tokens were minted at launch.
The network went live November 24, 2023.
For 9 months, miners ran compute and earned tokens.
With no market to sell them on.
When trading opened in August 2024, it opened on Uniswap. Same price. Same access. For everyone.
In a space where "fair launch" usually means
"we told our friends first", TIG opened on a DEX with no whitelist.
Now read the full on-chain breakdown.
Every claim is verifiable yourself.
@tigfoundation
On-Chain Investigation
THE TIG TRUTH
VERIFIED FROM BASESCAN Β· SABLIER Β· TIG DOCS
$TIG / BASE NETWORK
CONTRACT VERIFIED MIT LICENSE
A thread nobody has put together on $TIG yet.
I went on-chain. I read the contract. I pulled the data.
Here's what I found β and why it changes how you should think about this token.
The token contract has no genesis mint.
Zero tokens were minted at launch. No presale batch. No founder allocation sitting in a wallet on day one.
The mint() function is callable only by the protocol owner β and it fires per block, per round, as rewards are earned.
Contract Verification
CONTRACT0x0C03Ce27...9F7B
VERIFIEDβ EXACT MATCH
GENESIS MINTNONE
MINT ACCESSPROTOCOL ONLY
LICENSEMIT (OPEN SOURCE)
The network launched November 24, 2023.
The token hit Uniswap on Base in August 2024.
That's 9 months of the network running with tokens being emitted β and no market to sell them on.
Anyone benchmarking in those early rounds was working for tokens worth $0 on open markets.
Timeline
NETWORK LIVE24 NOV 2023
TRANCHE 1 EMISSION100 TIG / BLOCK
TRANCHE 1 END01 JUNE 2024
UNISWAP LISTING13 AUG 2024
PRE-MARKET PERIOD~9 MONTHS
No VC round. No seed. No named investors. Ever.
There is no public announcement, no Crunchbase entry, no named fund that received a discounted allocation.
When trading opened, it opened on Uniswap β not a launchpad, not a CEX listing event with KYC waitlists. Same price. Same access. For everyone.
In a space where "fair launch" usually means "we told our friends first" β TIG opened on a DEX with no whitelist.
The largest token holder on-chain is a Sablier vesting contract.
Sablier locks tokens into a smart contract schedule that no one can override.
~20 million TIG β roughly 40% of all currently minted tokens β is locked. Publicly. Verifiably. On-chain.
Sablier Vesting Contract
ADDRESS0x4cb16d41...78d8
TIG LOCKED~20,000,000 TIG
% OF MINTED SUPPLY~40%
OVERRIDABLE?NO β SMART CONTRACT
RANK#1 HOLDER ON BASESCAN
How did the team get their tokens?
The same way every benchmarker did. By running compute during those 9 months of zero-liquidity mining.
They then chose to voluntarily lock their stack into Sablier β a time-lock anyone can see, verify, and audit right now on Basescan.
There are also 275,001 TIG burned on-chain as of Oct 2024. Someone is burning. Not hoarding.
Supply Snapshot
MAX SUPPLY131,040,000 TIG
MINTED SO FAR~49,400,000 TIG
CIRCULATING~29,000,000 TIG
BURNED275,001 TIG
LOCKED IN SABLIER~20,000,000 TIG
Verdict
This isn't a marketing claim. This is what the chain says.
No genesis mint β tokens earned per block, not pre-allocated
No VC round β no named investor, no discounted presale, ever announced
9 months of zero-liquidity mining before anyone could sell
Team tokens locked in Sablier β publicly verifiable, non-overridable
~275K TIG burned β team is destroying, not accumulating
MIT licensed, open source contract β anyone can read it
VERIFY YOURSELF
Contract: 0x0C03Ce270B4826Ec62e7DD007f0B716068639F7B
Sablier streams: app.sablier.com β search 0x4cb16d4153123a74bc724d161050959754f378d8 on Base
TIG // BASE
ON-CHAIN VERIFIED
tig.foundation
DYOR β NOT FINANCIAL ADVICE
FhtAbd retweeted
If youβve been following $TIG for a while, you know that @Dr_JohnFletcher has been making these exact same points for YEARS.
Now, every 48 hours, another Fortune 500 CEO arrives at the same conclusions, admitting what weβve known all along.
Algorithms, not compute, will decide who wins the AI race and @tigfoundation is the most efficient means available for optimizing, monetizing, and proliferating them. It is purpose-built for this exact moment.
It will be worth billions because it has achieved generational positioning and product-market-fit. This is rare in general and unheard of in crypto. It will succeed for the good of the species.
Oh, and did I mention thereβs no competition?
FEW.
Jensen Huang (@nvidia CEO) thinks algorithms are the lever and open source wins long-term.
So do we.
Worth listening to this whole clip.
FhtAbd retweeted
Yep.. Prometheus from @tigfoundation is an absolute monster
$TIG
FhtAbd retweeted
So Prometheus is going to track the process of getting an answer and store it locally.
There's a useful distinction between receiving an answer and knowing how you get there.
You realize how important this will be for algorithmic research and where this will take $TIG?
It sounds like you'll be able to request information from other agents and pay (in $TIG?) for help too?
Damn. Getting more excited about this. Based on the work they already have out, I'm expecting something elegant and very well thought out.
I could point out something else in this tweet set, but I'll let you infer what you will on your own. Early.
Summary of yesterday's community call.
@Dr_JohnFletcher walked through Prometheus, its importance, and what it means for the next 25 years of AI.
