Hacker News Reader

tgtweakon An update on Wayback Machine access

Can't wayback machine just offer direct access to the archive for a premium and in doing so, pay for the service?

tgtweakon Patterns and problems in emerging multi-agent systems

Are we surprised? Humans evolved with communal success and collaboration engrained over millennia. Agents are trained as individual "all knowing" single entities, effectively rendering them single person players.

These models all have the same knowledgebase as well and thus see no value in the opposing agents contributions since they are "obvious".

Overall amusing but kind of expected.

tgtweakon 2026 Eclipse Webcams

https://eclipsefan.org/

Posting this from a HN post a few days ago - one of the better "where is the eclipse currently" trackers that isn't currently ddossed.

adding that there are some great webcams in santander on youtube that you can watch as it starts in a few minutes:

https://www.youtube.com/@tendnetworks

tgtweakon Mars Bar from 1991 found – and it's 20g bigger than today's

Check out the post from the family that has been counting Quality Street tin candies every year for the last decade or two... very telling.

tgtweakon 50k Boat Names

It's like vanity license plates but with even less constraints... "Unsinkable II".

tgtweakon Muse Glimmer: 30B-parameter model optimized for always-on local agent workflows

There is definitely an under-served target memory size of 48GB - almost everything aims for: 12, 16, 24, 32, 64, ...) But most dual-gpu setups, 3090/4090 (and some mac configs afaik) have 48GB, and most 64GB systems would do well with the extra 16gb of overhead saved. 48GB is also moderately common in PC memory configurations since 24gb DIMMs are a thing.

tgtweakon Launch HN: ProvenMetal (YC S26) delivers circuit boards in days instead of weeks

Amazing, now if we could just get this for silicon wafers :D All joking aside, would be great to get something like this for shuttle run wafer taping vs the current 6+ month window.

tgtweakon Mario Meets Pareto

2020: https://www.civisanalytics.com/resources/the-best-mario-kart...

OG credit to: Henry Hinnefeld 2015: https://hinnefe2.github.io/python/tools/2015/09/21/mario-kar...

tgtweakon The coolest use for the Vision Pro

Quest 3 has a 3d explorer (hyperscape) and you can walk around Gordon Ramsay's kitchen.. it was the first "ok, damn" app experience in the quest for me where it felt like you were there.

I can only imagine how powerful that experience would be for experiencing your future house.

Throw in some time-of-day lighting and sun-path illumination... would he great!

tgtweakon Hybrid-Electric Aicraft Engine Targeting 30% Fuel Efficiency

Says that it's boost-only in all the literature I could see, so it can only add power to the prop not generate with it. Regardless, you don't really get into a position where you're harvesting energy in a plane - you just use less power while you're descending. Unlike in a car, most of a passenger jet's flight time is at speeds where drag (which squares with speed) basically means you'd have to nose down at a very aggressive angle to actually pick up speed without the engines providing thrust. The plane's engines are almost always under some kind of load until it is on the tarmac and slowing down so there isn't any opportunity to "regen" during a normal flight.

tgtweakon Hybrid-Electric Aicraft Engine Targeting 30% Fuel Efficiency

More like an Edison Motors hybrid system.

tgtweakon Hybrid-Electric Aicraft Engine Targeting 30% Fuel Efficiency

Yes but then you're erasing some of your fuel efficiency gains on take-off by running a significantly less efficient APU to charge it back up. Also I think the APU is a built in unit designed entirely for in-flight loads (like an alternator on a car) and not intended to provide the kind of energy you'd need to even moderately recharge this pack after takeoff/ascent.

tgtweakon Hybrid-Electric Aicraft Engine Targeting 30% Fuel Efficiency

The larger turbine (like a PW100 1.8MW in the dash-8 that this demonstrator is replacing) is at peak efficiency (~0.30kg/kWh output SFC) near full load, then it cuts back to 50% power while cruising where it also drops into a less-efficient SFC rate (~0.36) then down again to 20% (~0.45) for descent vs the "always at 100%" 1MW version which stays pretty much pinned at 0.30kg/kWh sfc sweet spot during all of the flight except descent where it also drops back and takes an efficiency hit.

tgtweakon Hybrid-Electric Aicraft Engine Targeting 30% Fuel Efficiency

Rtx has an interesting patent [1] on this that highlights some of the novelties of this setup vs a traditional hybrid (planetary motor/generator like in a Prius):

It's a boost-only motor, it doesn't/can't harvest energy on descent.

The patented solution (transient smoothing under auto-throttle control) puts electric motors on both the low spool and high spool, then uses a power-splitting algorithm to route high-frequency thrust changes to the electric motors while keeping fuel flow nearly constant on the thermal engine (turbine). The turbine cruises at a steady operating point with tight compressor/turbine clearances and the electric motor smooths out the spikes that are normally there with turbulence and load changes. Benefits: lower fuel burn, longer turbine life (fewer blade-rub risks from speed variation), and smoother ride quality since the auto-throttle bandwidth improves. This setup (based on the various cutaways and photos so far) seems to be only a single 1MW motor so it only runs on the low spool but can still help modulate the turbine decently in the same way it does in the Koenigsegg Regera's hybrid electric setup, that removes the need for a flywheel because the electric motor can smooth out the gas motor's inherent lumpiness.

Also disclosed in a previous press release [2], it's only a 200kWh battery so at 1MW peak boost (cited load during takeoff/ascent) it would only run for ~10-15 minutes at the beginning of the flight.

Seems most of the savings are due in part to not using as much fuel during takeoff (~20% of a 1-hour flight's fuel) but also in large part to the under-sizing and optimization of the thermal turbine to keep it running in it's peak efficiency zone for more of the flight (~10% of a 1-hour flight's fuel).

Curious how the safety margins work here - if the battery is depleted on takeoff (aborted takeoff) or there's an issue that requires descent-then-reascent, if the batteries can't be replenished in-flight there could be a power deficit in that window where you'd normally have 2+2MW of gas turbine power for the plane and now you only have 1+1mw of gas turbine power.

[1] https://patents.google.com/patent/US20250296689A1/en

[2] https://www.aerospacetestinginternational.com/news/h55-deliv...

tgtweakon OnePlus halts operations in USA and Europe

The cameras on the Pro versions used to be so good, then starting around one-plus 9 they really went downhill. I have photos in my archive from my OP8 Pro that look much better than those taken with the OP13.

Coinciding with Samsung nerfing the Ultra (aside from the bloatware) - it's not looking like a great landscape for Android Phones.

tgtweakon Kimi K3: Open Frontier Intelligence

I think given how much benchmaxxing we're seeing - the anecdotal evidence of how competent this model is (and efficient) will depend on user's actual real-world use cases.

Given the pricing, it suggests that this model is much more efficient/competent than previous-gen OS/distilled models.

tgtweakon Claude Code is steganographically marking requests

None of this is surprising - they're trying to mask and relay when they detect known patterns of what looks like distillation attacks and client app copying/modification. The list obfuscation here is likely to prevent or make it difficult for those same adversaries to work around this or delete/null it out when making a bootleg copy.

Cool reverse engineering/analysis report but if this is the extent of nefarious activity that came of it (trying to catch/mitigate chinese lab model distillations), that's kind of encouraging.

tgtweakon FUTO Swipe – A new swipe typing model

I feel like a clean room opensource keyboard could implement those without any patent liabilities. Software patents are notoriously difficult to enforce and I don't think nuance is spending any money enforcing this given they stopped swype in 2018.

tgtweakon FUTO Swipe – A new swipe typing model

swype was so good before it got destroyed by Nuance. Gboard's own swipe to type is decent but missing a lot of those features that made swype so much better (swipe once backwards to erase the previous word, loop a letter to double it, swipe up to the suggested word to select it, correct a previous word with the next few words context...)

tgtweakon Polymarket has flooded social media with deceptive videos by paid creators

It's the same for all the gambling platforms - they give creators and influencers privileged accounts that win more than they lose so when they stream it looks like they're not losing constantly.

tgtweakon Canada plans 'nuclear renaissance' with up to 10 reactors built by 2040

Nuclear also works well with grid batteries to smooth demand curves, which Ontario is targeting 2700MW of scale by 2030.

tgtweakon Codex logging bug may write TBs to local SSDs

Not a decision that users are aware of though. Nor is there a setting to disable/change it. It just showed up one day and erased your previous sessions.

tgtweakon Codex logging bug may write TBs to local SSDs

Slightly better than the claude code "feature" that deletes all your session context and transcripts older than 30 days old.

tgtweakon I found 10k GitHub repositories distributing Trojan malware

This is a failure of malware flagging systems as well - VT should not return clean if there are any downstream files that are malicious - such as in this case.

tgtweakon Claude Fable 5

Implying that google's "snippets" were never curated to remove anti-google facts or that they didn't curate the search results in their favor...

tgtweakon Building an HTML-first site doubled our users overnight

It'll be replaced by a new react app within a few hires lol

tgtweakon Gemma 4 12B: A unified, encoder-free multimodal model

It's a disruption game - releasing competent open models disrupts smaller labs trying to release their own or commercialize their own. It's a similar rationale behind the Chinese labs releasing near-frontier open-weighted models, the goal is to disrupt and lift the barrier of entry for would-be competitors.

tgtweakon Use your Nvidia GPU's VRAM as swap space on Linux

I think you can definitely improve the throughput/iops by using BAR vs treating it like a file store/mount through cuda which adds a lot of overhead.

tgtweakon Shopify Is Down

Critically, it was the webhook/sync that was down which really messed with a lot of external systems (nosto, klaviyo, 3PLs...)

tgtweakon MAI-Code-1-Flash

Is anyone using haiku 4.5?

Why not showcase it against something in a similar domain like qwen3.6 or gemma 4?