Promise.all isn't a concurrency limit.
12 local timer jobs:
.map(task) + Promise.all → peak 12 in flight
3-worker queue → peak 3
Both finished all 12.
For batch API work, limit how many jobs you start. Promise.all only waits for the results.
hey @typesafeai I think your token usage page has broken
it was working till yesterday
and now only the balance is going down
the token usage is still showing yesterday's data
so I gave jev full control of the falcon-9 class rocket
full launch to landing
no autopilot, no scripted trajectories, no safety veto in the code
every decision was made by jev ( you can see decisions at the bottom of the video)
and yeah it was able to land the booster
and all this with just 245 calls and and $0.04 in API calls
for sim I used mujoco
and the configurations were 42m booster, 425 tonnes, 9 engines, real atmosphere, mach dependent drag, wind and propellent depletion
the vehicle was going ubstable every .8 secs, and jev's decesion making takes place at ~1Hz and output in ~0.4 sec
so I gave jev these controls
> launch or hold
>engine count for every burn( it chose 5 for landing by itself !)
> when to light that burn
>ascent cutoff, coast, entry, legs, touchdown
so after giving all these controls to jev the experiments begun
and after 12 runs and tweaking multiple things it managed to land by itself
I will write a full article about it
and what I did with experiments and whet I tweaks I made
but for some final thoughts
people are saying it's just a classifier
yes, but it's a very good and effective classifier, and its speed and decision making, open up a whole new work for real life tasks and implementationa
I will also open source the whole repo for this after cleaning up the clutter in my code
so this is the short demo
I hope you like it
day 3 update of Fly MMA live streaming
live streaming platform is almost done
trained ~54k new training decisions on my M1 and GTX 1660 Ti
with 165,122 simulated neurons per fighter
Added live commentary and chat and
fly audience can now bet on the matches as well (250 flies spectators)
other people will also be able to submit their fighter flies to match up with the current fighters
next up
> deploy this on public domain
> do live MMA stream on Youtube
quick update on the MMA fly
we got ~15k views in 1 hours of it's launch
full training also finished total 1062 training episodes
now flies can do complex maneuvers and fight with different techniques
also this matchup was proudly sponsored by @raykolabs
other sponsor spots are still open DM me for the spot
next up
>training for spectators is still going on we're going to have around 250 spectators who will comtinously bet on the fly fighters live
> live stream fly MMA 24/7
so I trained my fly to be a MMA fighter 🪰🥊
actually I trained 4 flies out of them I chose 2 to fight each other
165,122 simulated neurons.
and also the moves are not fixed
they choose their own moves according to the opponent
and winner gets the champion belt
and it also did a headstand on winning 😂
next up
> training spectator flies to actually react to the match
>and also teaching them to bet on the fighter
>live streaming fly MMA 24/7
so I have been running multiple experiments with fly brain
in shorter training experiments it is able to learn short actions very easily
but in longer rung I still need to find out the progress and if it able to perform autonomous actions instead if just learning predefined actions
let's see how it goes
I will keep posting the updates
the SS is just an html render generated with astra to show the training progress
I have added deepseek benchmark comparison with GPT-6 astra and fable 5.1
and to my surprise it holds pretty well
at 98% lower the cost which is 70-80x less the price of frontier models
If anyone wants to understand why Navier-Stokes equations is such a big deal and what makes openai cracking it even bigger
I have made this interactive visualization so that anyone can understand it without any mathematical technical knowledge
please give it a read
hope you like it
Are websites dead?
Feels like nobody browses websites anymore.
People just ask AI, search on TikTok/Instagram, or land directly from Google on one page and leave.
Most websites today feel more like trust pages than discovery platforms.
im officially a blog partner of shipsquad
shipsquad is not just another product listing site
it helps builders turn early supporters into promoters who drive growth
the next big thing might just be one scroll away
New indie launches are up!
From scrappy experiments to polished tools, all made by builders chasing real problems.
Support early, discover hidden gems, and maybe stumble on your next daily driver.
See what’s new 👇👇
Linkto @RafVantongerloo
JustPageit
HypeDesk
Chart AI - Technical Analysis @ahbeosi
Reality Check
GumLeads @GumLeads
Date Everything
Feednap @Feednap
Random Animal Generator
LinguaLoop @LinguaLoopApp
TypeThinkAI @YashPat51114437
Archetiq
ForgeBoard @BenBoarer
Billr @MOulitzky
MutualGro @mikel_farrell
CloudTrain
Zalvis @zalviscloud
RunAgent
Feednap @Feednap
ObfuscatorJS
Pick Who Pays
SummarizerDM
Random Game Generator
ArrayCat
PicMotionAI @aka_govind
Grow a Garden Calculator
Crosspost
Auxigent
Turi
YakStak app
Explore and show the builders a little love today 💛
With AI booming and India's potential to create innovative AI tools, I noticed a gap:
- No single free platform exists for AI enthusiasts to
- Discover, discuss, and support new AI tools, especially from indie creators.
So, I built AI Exchange, a free, open platform