Lisa Collins

Experience Platforms Director @Accenture; views my own.

[Taelin @VictorTaelin]

this is my personal singularity moment

this post may sound like a paid ad. I only wish. I'm concerned, more so than happy. the world is changing, and, among the scenarios where AI goes terribly wrong, inequality is the most realistic, yet, the one Anthropic seems to be the least concerned about. I'm glad OpenAI is taking the opposite stance: *personal AGI for everyone*. I think this is a commendable position in the times we live. but who am I in the queue of the bread?

anyway, Fable is here, so I'll just report my first-hour experience

first of all, all my pet prompts are solved.
→ λ-calculus puzzles
→ bug questions
→ one-shot apps
all are trivial to it.

I don't have anything harder other than my
ongoing work

so, in the last several days, I've been toying with HVM5, a new interaction net evaluator with a faster loop.

after writing the first version, I left 32 GPT-5 agents working for ~20 hours each. this resulted in up to 2x speedups, but the file size increased by 2-fold and quality decreased significantly.

I then simplified the whole thing into an even simpler core, and left Opus 4.8 and GPT 5.5 optimizing it for 8 hours. Opus got a legit 6% - 34% speedup in most benches. GPT got better results, but, sadly, an unusable file.

I then asked Fable to optimize it.

2 hours later, it landed a 1770% speedup in one case, 100%+ in other 4, and 22% in average. yes, in 2 hours it outperformed me, opus 4.8 and a swarm of gpt 5.5 agents, by one order of magnitude.

that could not possibly be legit. "it must be hardcoding the benchmarks" (GPT trauma). so I read its explanation and what it did was, indeed, the most high impact optimization one could try first. seems like HVM5 was wasting a lot of time garbage-collecting unused branches of pattern-match nodes. I had optimized that for static mats, but not for dynamic mats. skill issue. Fable figured how to do it for these, resulting in a massive speedup in some benches

but wait, is that *correct*? I'm not sure yet, it is credible, but this is the kind of thing that is very easy to get wrong on interaction nets. the problem is, when I was ready to start auditing Fable's solution so I could tell whether it was buggy or legit, it interrupted me to tell me it had found a massive bug on the code *I* had written.

... wait, what?

so... for garbage collection purposes, I stored a bit on lambda term pointers that meant "the variable bound by this lambda has been freed, so, its lambda must free whatever argument it is applied to". that's fine. yet, on duplicator nodes, I also used the same bit to mean "one of the duplicated variables was freed, so, treat this dup as a passthrough no-op". so, if a lambda entered a duplicator, it would mistake the lambda's collection bit for its own, resulting in corrupted interaction!

that's a mouthful, why I'm writing this?

just so you can appreciate the sheer absurdity of what just happened. I didn't ask it to find bugs. I asked it for an optimization. and even if I did ask it to find bugs, this bug is so astonishingly subtle and specific, identifying it takes mastering the domain to an extent that it beyond even me. I'd easily need hours or days to fix it, *if* I ever came across it. chances are it would just go unnoticed. and Fable found it and fixed it like it was nothing, while it was busy adding a 17x speedup to a file that neither I, nor Opus 4.8, nor a fleet of GPT 5.5 managed to barely make 2x faster.

oh and there is also another tab where it is also ripping through Bend's codebase and finishing everything I had to do

I don't know what to say anymore

this isn't about Anthropic or OpenAI, this is about our collective future as a species. the world is changing, and we need to be aware of it, and discuss how to handle this change.

receipt below . . .

[Lisa Collins @LisaCollinsSG]

@dimiwonders Wonda

[Lisa Collins @LisaCollinsSG]

Out of context tTms call quotes: "My RAG isn't going to look like someone else's RAG".

I've been saying the same thing for years.

[Brendan Jowett @jowettbrendan]

🚨 BREAKING: Google just launched a new internet protocol.

It’s A2A.

Backed by 50+ partners including Salesforce, Atlassian, and SAP.

It’s designed to let AI agents collaborate across companies, platforms, and clouds.

Here’s what it means for the future of enterprise AI: 🧵

[Bim | Lukas @bimXBB]

for anyone interested in blockchain analytics and catching onchain crime

@zachxbt just published a huge list of tools he uses

[Lisa Collins @LisaCollinsSG]

Handy reference comparing different LLM models (real-time): https://t.co/bzgJ7IHcAA

[Allie K. Miller @alliekmiller]

NEW AI products that do this since my tweet two weeks ago:

https://t.co/jFmg4yxCMH now provides a full-McKinsey-like report through its Ari tool. Ex: https://t.co/zXRRxUTSQS

Manus AI agent providing a deployed website to view a research dashboard. Ex: https://t.co/etUzYCUXxQ

[Allie K. Miller @alliekmiller]

The AI skills gap isn't about coding or prompt engineering—it's about systems thinking.

Companies scrambling to hire LLM experts are missing the point. We need people who understand both the business process and how AI can transform it.

[Lisa Collins @LisaCollinsSG]

@SpotifyCares hi where do we report abusive (borderline?) illegal content currently on Spotify?

[Justin Wolfers @JustinWolfers]

And here's the number of RECIPIENTS of social security in each age bucket with the death field set to false (and recipient set to true). A mere 89,106 are aged 99+, not the tens of millions suggested by @elonmusk.

[Lisa Collins @LisaCollinsSG]

Long but worth it - without a doubt the most fascinating take on DeepSeek (and Nvidia) to emerge last week:

[Jeffrey Emanuel @doodlestein]

You can read it here— let me know what you think:

https://t.co/NjBBMX7nS1

[Lisa Collins @LisaCollinsSG]

Today I have a deep desire to leave the city and to go be in nature for a month.

[DatNoFact ↗ (@datnofact.bsky.social) @datnofact]

hello I'm new to the stock market is it good when the intel ceo starts praying

[Lisa Collins @LisaCollinsSG]

What a surprise to find such a gem from the son of Estée Lauder, Leonard. From his involvement in the biz from an early age, how he fought the Revlon/Lancôme wars, and the evolution of his leadership style and strategy was told with warmth and skill.

https://t.co/RsptPaBLUk

[Sam Whyte @samwhyte]

Whoever's responsible for the Microsoft outage is getting fired anyway, so the smart thing to do would be knock Teams out for the day too and leave a hero.

[Michael Friendly @datavisfriendly.bsky.social @datavisFriendly]

Excellent paper 💯
I'll probably end up stealing some of the images/illustrations (in the best sense of stealing)

[Nick St. Pierre @nickfloats]

Midjourney just released a new feature called 'model personalization'

It lets you tune the MJ algorithm to your own personal tastes, removing much of the MJ "bias" that comes from its training data

Breakdown of how it works:

[Nicolas | Gallet Capital | The Trading Alpinist @TradingAlpinist]

#Tradeidea fur June 2024: generate over 50% APR building a Synthetic USD position using @ipor_io (very low risk quasi delta neutral position) using the wstETH mining function powered by IPOR staking on Arbitrum

As I am about to take my fat summer holiday, I am looking at diversifying my delta neutral positions for easy, safe and high APR while not looking much at the screens.
And it is raining yields currently on IPOR.
Liquidity mining on IPOR V2 using Arbitrum currently provides amazing opportunities, boosted by IPOR's LTIPP summer campaign distributing 112,500 $ARB, Lido incentives of 4 $wstETH and 142,000 $pwIPOR.

The trade I am about to describe is similar to Ethena synthetic USD in its logic but the return is much higher and most importantly certain, it embeds a great upside for even higher return, and importantly the centralization risk divided by 4 in comparison to Ethena. In more details, in this trade you will earn cumulatively :
the power of IPOR staking,
the ETH staking rate,
ETH contango rate,
LTIPP ARB incentive
Lido wstETH incentive
pwIPOR incentives

This is for education purpose only, I will make it simple, up to you to decide if you are a Ethena crowd moron or a smart DeFi user.

My current portfolio of passive yield generation includes Pendle PT positions on uniETH and rzETH, Spot vs dated futures basis, opportunistically spot vs perps in GMX (synthetic decentralized USD) and the whole portfolio currently runs at around 43% APR daily. In short the bar is high for diversification, but this trade idea fits the league and will likely stay in the portfolio for the coming year.

The problem: finding the right allocation of $IPOR to delegate against a $ETH staked position (zapped in $wstETH )and hedged by $ETH dated futures to enjoy a delta neutral position on ETH and an extra premium on the basis between spot and dated future.

I have computed the best capital allocation that minimize volatility of the portfolio.
my finding for a capital of $10k
$2310 or 23.1 % in $IPOR to be staked in delegated pwIPOR
$5380 or 53.8% in $ETH to be zapped in staked ipwstETH
$2310 or 23.1 % in $ETH to place as collateral in Deribit
a short position of $7690 in 26jul deribit future contract

in another words, a portfolio composed of 23% IPOR and 77% synthetic USD for at least 50%APR

You do not need to go to a staking platform to generate your wstETH using your ETH, IPOR does it for you using their Zapping function. And that s a blessing, as the whole trade would be tedious to put in place on your own. I have just adjusted the cursor on their dashboard for the desired USD equivalent balances of ipwstETH and pwIPOR after zapping.

Obviously a directional position on $IPOR is inevitable, but
1/each IPOR upgrade is a major support for its ecosystem, and IPOR FUSION looks like a major disruptive innovation in the Asset management space.
2/ when your capital allocation in IPOR is 23% while the APR is over 50% on your capital, it does not really matter where $IPOR is going, I will still secure a fat return even if $IPOR underperform. Nevertheless, optimizing the risk/rewards of such a position still makes sense. The volatility of the trade is diluted enough, such as using a z-score model, the probability to lose money from the direction of $IPOR over the coming year is near zero.
3/ The beauty of the IPOR tokenomics lies in the virtuous circle embedded within the model. While I do not claim to possess the same brilliance as the minds behind it, my implementation of numerous trades based on empirical modeling has shown that the proportions are usually similar, with 20 to 25% in IPOR tokens supporting a delta-neutral position. This low percentage effectively dilutes the risk of the entire high-yield strategy, significantly enhancing the token's utility in generating such yields. Thus, the model inherently supports a virtuous circle.

To create the synthetic USD position, I have neutralized my $ETH exposure by selling futures value 26july on Deribit for the same ETH notional with a x3 leverage. The contango (difference between future price and spot price) is really amazing right now at 16% APR without leverage. A more conservative leverage would not affect the potential of the trade, but x3 leverage for such a shot tenor on futures is fine, i expect in fact this spread to widen more in the short run, so I will probably roll against the next futures dates opportunistically later on. [ note: When @cvex_xyz will go live, I will be able to do the whole strategy fully in DeFi, avoiding the CEX risk. The Synthetic USD will be fully decentralized at that point, i will not rely on a Bahamas or Seychelles moronic traders to sit on those massive derivatives positions praying they understand what risk and collaterals mean, it will be finally programmatic]

I optimized the proportions of my engaged capital (IPOR + stETH + collateral on Deribit) against the volatility of such a portfolio. As usual I use annualized ATR as a less conservative estimation of real volatility. Generating roughly 50+% with a return/risk ratio of 3.3
I am probably a bit aggressive on my weights ratio, emphasizing on the virtuous effect of the IPOR tokenomics. A safer statistical bet (higher risk reward ratio) would eventually favor a lower proportion of IPOR in your combo (17% of IPOR allocation for risk reward of 4.25 and minimum return of 48%). I attached my model results to adjust the IPOR allocation to ones risk appetite
I made the assumption that 25% of the ARB incentive and 100% of lido incentive will be dedicated to that pool. And still I did not factor in my computation : the pwIPOR inventive as an extra bonus on top, and the few extra percent from ETH staking rates. 50% looks like a base case

As usual, get your parents guidance and surveillance

#APR #passiveyield #IPOR #liquiditymining #Staking #iporfusion #wstETH
If you like this post, please share, follow, comment, retweet, DM

[Vogue Business @voguebusiness]

Balenciaga ranked third overall with $3.8 million in EMV by looking beyond established names. Rising Thai star, Beccca and her three impactful posts for the brand garnered a remarkable $3.69 million EMV, or 75% of the total EMV for Balenciaga. https://t.co/dmE6HD8uQY

[Alex Reibman 🖇️ @AlexReibman]

Mistral is France’s answer to OpenAI. And it’s all open source.

They just threw Paris’s largest ever AI hackathon. 1,000+ hackers applied to build what’s possible with open source LLMs.

Here are the finalists from the @MistralAI x @cerebral_valley hackathon in Paris (🧵):