arkhet.hl
1.5K posts

arkhet.hl
@arkhet
ad astra 💫 @chutes_ai evangelist 🪂 composability is the future 🧱
🇺🇸🦅 Katılım Ekim 2024
3.2K Takip Edilen3K Takipçiler
Sabitlenmiş Tweet

I think there's an underrated aspect of investing in Bittensor.
In other ecosystems, you are competing with an infinite number of projects/tokens. On Bittensor, the current competition is among 128 subnets.
Out of those 128:
<3 are elite
10-20 are doing good job
The rest is either super early/highly speculative or straight scam (easy to spot).
So your job is not that complicated.
Thesis remains: Value will concentrate on top subnets, it is a matter of time.
English

@ImtiazMadmood this is upsetting ☹️
not compatible with western culture
English

Most prominent terrorist attacks since 1970 and their religion:
1. Munich Olympics massacre (1972) - Islam
2. Beirut barracks bombing (1983) - Islam
3. TWA Flight 847 (1985) - Islam
4. Rome & Vienna airport attacks (1985) - Islam
5. Pan Am Flight 103 (1988) - Islam
6. World Trade Center bombing (1993) - Islam
7. Paris Metro bombings (1998) - Islam
8. US Embassy bombings Kenya & Tanzania (1998) - Islam
9. USS Cole bombing (2000) - Islam
10. 9/11 attacks (2001) - Islam
11. Bali bombings (2002) - Islam
12. Istanbul bombings (2003) - Islam
13. Madrid train bombings (2004) - Islam
14. London 7/7 bombings (2005) - Islam
15. Fort Hood shooting (2009) - Islam
16. Toulouse/Montauban shootings (2012) - Islam
17. Boston Marathon bombing (2013) - Islam
18. Brussels Jewish Museum shooting (2014) - Islam
19. Ottawa Parliament attack (2014) - Islam
20. Charlie Hebdo / Hyper Cacher attacks (2015) - Islam
21. Paris November attacks (2015) - Islam
22. San Bernardino shooting (2015) - Islam
23. Brussels airport/metro bombings (2016) - Islam
24. Nice truck attack (2016) - Islam
25. Berlin Christmas Market attack (2016) - Islam
26. Orlando Pulse nightclub shooting (2016) - Islam
27. Westminster Bridge attack (2017) - Islam
28. Manchester Arena bombing (2017) - Islam
29. London Bridge/Borough Market attack (2017) - Islam
30. Barcelona/Cambrils attacks (2017) - Islam
31. Strasbourg Christmas market attack (2018) - Islam
32. London Bridge stabbing (2019) - Islam
33. Samuel apart beheading (2020) - Islam
34. Vienna shooting (2020) - Islam
35. October 7 Hamas-led attacks (2023) - Islam
36. Arras school stabbing (2023) - Islam
37. Solingen knife attack (2024) - Islam
38. New Orleans Bourbon Street attack (2025) - Islam
39. Bondi Beach Hanukkah attack (2025) - Islam
40. Winterthur train-station stabbing (2026) - Islam
I couldn’t list them all, obviously, but let’s put it into perspective.
Islamist terrorist attacks worldwide since 1979:
1979–April 2024: 66,872 attacks
By period:
1979–2000: 2,194 attacks
2001–2012: 8,265 attacks
2013–April 2024: 56,413 attacks
Total Deaths: at least 249,941
Now, who is the biggest threat to the world, again?
English

First-order: @LightmatterCo tech makes chip-to-chip communication much faster and way more power-efficient.
Second-order: Big AI companies can now build larger, more efficient training clusters. Training gets faster and cheaper.
Third-order: Inference becomes cheaper too; same hardware can now run models with lower power and better performance.
Fourth-order: Decentralized networks like Chutes AI benefit. More GPUs get freed up, hardware becomes more affordable over time, and inference margins improve for miners.
End user benefits in ttft, quicker inference overall, plus increased frequency of model refinements/improvements
arkhet.hl@arkhet
@chutes_ai will be a 3rd or 4th order beneficiary of this imo
English

i’ve had a strong conviction in photonics for some time now, but Nick Harris (@theanalognick) is killing it
very impressive execution from @LightmatterCo
arkhet.hl@arkhet
using the speed of light to our advantage youtu.be/8rO3dwrlSqk?si…
English




In a future iOS update, when you can Bring Your Own Key for Siri
What would you choose?
centralized, close source inference is a total violation of privacy - no way around this. Prompts and data are logged and trained upon.
decentralized, open source inference from providers like @chutes_ai cryptographically guarantee post quantum privacy.
There’s a route the sheep may take, but those who have a sense of awareness likely may choose the latter.
position yourself accordingly - the future is coming towards us at an ever increasing velocity.
English

Edge devices that cannot fit local models will inference decentralized intelligence @chutes_ai
Edge devices that can, wont be as efficient (weight, speed, power effeciency)
Unless you’re okay with camera, mic, telemetry, sensor data, etc. to be constantly streamed to any centralized counterpart
Parallax is Patent Pending 🪂
English

Robots and Physical AI will run on @chutes_ai decentralized intelligence
youtu.be/21BzAy5YEuE?si…

YouTube
English

Everyone's out here waiting for aliens to arrive.
WE are the aliens.
A camera we built is orbiting mars right now, staring into a crater no living thing has ever seen. crossed the void, showed up unannounced, peeked inside.
In this story we're the visitors from another world.
and honestly? kinda iconic of us 🪐
English

been away from bittensor:native for a few months? here's everything that actually mattered, jan to may 2026
january, institutional money started moving into TAO for real. Grayscale filed, Bitwise filed, $620M+ came in from Nvidia and Polychain-level names. Upbit listing too and while all that was happening, builders didn't slow down. CrunchDAO brought 11k+ ML engineers on, subnet ideathon ran, SDK upgraded, coldkey security fixed. two completely different types of growth happening at the same time.
february was the one most people slept on. Jacob Steeves stepped down as OTF CEO and the network went fully headless. no foundation lead, just pure on-chain governance and community. stake burn, 256 UID expansion, locked stake conviction mechanics, dTAO flows picking up. the protocol stopped feeling like a startup and started feeling like actual infrastructure.
march was the proof drop everyone had been waiting for. Covenant-72B. 72B parameter model trained fully decentralized on consumer GPUs, beat Llama-2 70B on benchmarks, open weights released. TAO ran hard. skeptics genuinely had nothing to say for once.
april got ugly though. the same Covenant team that just delivered march's win did a full exit. subnets shut, open letter, allegations, 15-20% dump in a single day. market panicked. i didn't. because while all that drama played out, Q1 revenue came out at $43M, actual revenue not TVL, neuron registration rework went live quietly, and Grayscale bumped TAO to 43% weight in their AI fund. one team walking out didn't break what was already built. that's the whole point.
may, brutally competitive now. Dynamic TAO full force after the Taoflow upgrade, emissions tied purely to net staking inflows. capital leaving your subnet means zero emissions, no exceptions. subnet economy looks like hundreds of independent AI businesses fighting for capital in real time. Yuma + Crypto.com partnership, TaoWeave reports, core team AMA gave full clarity. weak stuff is getting filtered naturally. strong real use cases are winning. exactly how it's supposed to work.
what five months actually showed:
institutions aren't studying TAO anymore, they're buying it. builders shipped through a full blown crisis without stopping. and the subnet economy is self-sorting now, doesn't need any single team to survive.
been heavy in bittensor:native through all of it. conviction hasn't moved. this stretch honestly made me more sure, not less. the people who actually understand what's being built here will look back at this window very differently.
#Bittensor bittensor:native

English








