Hacker Newsnew | past | comments | ask | show | jobs | submit | ggregoire's commentslogin

I haven't been disappointed with the SF shows coming from Apple TV. Hopefully this is another great show.

Can you give some recommendations?

Silo is best imo. The upcoming season 4 is supposed to complete the story.

The first seasons of Dark Matter and Severance were great. Murderbot was good, but not great.


Great:

- Severance: psycho thriller in a time where people can surgically split their memories between work and personal life

- Foundation: space opera and politics, visually stunning

- Silo: dystopian thriller where people live in underground bunkers for mysterious reasons

- Pluribus: comedy/drama/psycho thriller where an alien virus takes control of every human on earth, except for Carol

Good:

- For All Manking: alternate history, space exploration, geopolitics (first two seasons are great)

- Dark Matter: drama thriller about the multiverse

- Murderbot: robot soldier that becomes self aware, comedy


   Meh.  Streaming services seem to do a thing where they put
   a  lot of  effort into  the first  season to  try to  lure
   people away from rival  services.  There's less motivation
   to make  subsequent series  decent.  I  sort of  hate what
   Amazon did  to The Expanse,  but we  did get at  least two
   decent seasons out of them.   All the series you list here
   were MUCH too  slow to maintain my interest.   I will give
   Apple props  for making  second seasons that  were decent.
   Third and subsequent seasons probably should have not been
   made.  But... if you like  them... you like them.  I'm not
   going to  argue you shouldn't.   I really liked  the first
   episode of Pluribus,  the first season of  For All Mankind
   and Murderbot though they were all very different things.

   I will argue  that a steady diet of MCU  movies has dulled
   our visual taste-buds.  As an experiment, try listening to
   just the audio track.  With the exception of Severance and
   Murderbot it's not hard to follow the story without seeing
   any of the action.  Film is supposed to be a visual medium
   using composition and motion to contextualize the words of
   the narrative.  Most of the items on the list fail in this
   regard.

   You can  argue that  Television is  a writer's  medium and
   that we  just have  to cope  with smaller  budgets, (MUCH)
   slower pacing  and creators  who are  more used  to having
   readers   create   their   own  visuals   in   their   own
   imaginations.  I think that's a cop out.  Guess that's why
   I'm not in the writers' guild.

   But... if you enjoy  something.  Don't apologize and don't
   let jerks like me diminish  your enjoyment.  For what they
   are, most  of these series  are entertaining.  I  just get
   distracted thinking about how  much better they could have
   been.

Silo is awesome

Sugar

Have you tried 4.1 flash? I'm using it to do machine learning in an audiovisual data pipeline, I basically do nothing, I only provide labeled data and it fine-tunes the algos, thresholds and so on in a loop, and publishes the evidences in markdown tables, creates images illustrating the clusters it found (I never asked for this!), all by itself. I was late and skeptical of the AI assisted coding but this is truly amazing.

And what absolutely crazy is that I used 2.4B tokens this month and it cost me $1.7 (w/ 99.4% cache hit) of my $10 sub


If you like 4.1 flash, what is your opinion on opus 5.5? Throw that at your pipeline. Just tell it to find the deepseek session (just naming the harness is enough, e.g. pi/opencode) and to proceed with the work. Then tell is the experience (genuinely curious). if you don't have a sub yet, 30USD/m will get you quite a lot of opus 5.5 I think nowadays. I used to have the 200USD, but now have the 100USD one, which I can still run out of with 24/7 agents but not with lighter work.

I use 4.1 flash pretty extensively for translating mandarin to English and it is noticeably worse than claude. In terms of tasks complete I find deepseek to be more expensive than the claude subscription I have as well. Deepseek uses an insane number of tokens per task.

Is there a similar project for React? There are so much bad React code and wrong/old design patterns on the internet that surely ended in the training datasets


a barely coherent ramble by a Microsoft engineer feat. an exquisite conflation of React/using libs like lodash or moment.js/the idea of frameworks, casually dropping false dichotomies between using React and delivering semantic HTML with progressive enhancement, drive-by smearing any use of frameworks by calling it "frameworkism" and equating it with religion, with a side of amusing bombshells like (exact quote)

> prohibiting the spread of React (and other frameworkist totems) by policy is an incredible cost savings and a helpful way to reorient teams towards delivery for users

isn't "simple".



Are they? (genuine question, i've got no experience with Claude) This month I used 1 billion tokens (including cache) with DS4.1 flash and it cost me 7% of my $10 opencode go sub.

If you meant in developer time, then I don't know but it seems very fast to me (rarely thinks more than ~10sec, never more than ~30sec) and usually oneshots what I prompt it to do. I use opencode tui as harness.


In my experience yes. Been using Claude since day one and wanted to experiment with Pi for fun, and try to get my costs down. I found the open models more expensive, eventually had to plug my Pi set up into an OpenAI subscription, that was the only way to get good value… Then I realised OpenAI is way behind Opus 5.5 and their subscription is far less generous with usage, so I’ve come full circle and am back on Claude Code.

i found they were until recently. mimo 2.6 flash and glm 5.3 flash seem to finally be cheaper. i find ds4.1 flash still on the high side.

I found GLM Flash competitive price wise but then my tasks took twice as long to complete. Even if it was cheaper (in my experience it still wasn’t) you then lose out on speed because of the insane amounts of inefficient reasoning.

Opencode go ($10), deepseek 4.1 flash high, opencode tui

I went directly with this setup, never tried claude/codex/cursor and their respective models.


When I order a Uber I just want a BYD… They should add the option in the app. Using the "Comfort" option increases the chance to get one, but I still get very bad/old cars occasionally.

Or maybe BYD could just start its own ridesharing service.


I'm sure a lot of people tried 1 model/harness and stuck with it without really comparing (same way I haven't changed my IDE and my terminal for the past 10 years), and eventually will try a Chinese model and be blown away.

I was discussing with a coworker yesterday who pays $200/month for cursor, has never tried any other harness and mostly use grok, in rare occasions opus 5.5. He seems to be regularly out of tokens. I told him I exclusively use DS 4.1 flash with opencode, spend like 100-150M tokens a day including cache and it oneshots all my tasks without issues. I barely reach 20% of my $10 sub at the end of the month. From his look I think he didn't believe me.


Meh… you can't create duplicated questions on SO so obviously 20 years later there is not much left to ask.

A better metric would be visits per month… which probably doesn't look very pretty too. I used to visit SO every day several times a day for like 15 years: ask a technical question on google, click on the first SO result, get the solution I was looking for. Now I still ask google but I just have to read the AI Overview to get the solution I'm looking for…


Yes, this can be abused to rewrite history too.

Related: Preserving Historical Truth: Detecting Historical Revisionism in Large Language Models

https://arxiv.org/html/2602.17433v1


"hacker" news… Even in the AI era, some people still enjoy understanding the technicalities of things

I think it was a great blog post.


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: