Fast charging has always carried a battery-life penalty. So BYD’s endurance test caught my attention. Its FLASH system is rated up to 1.5 MW, with 10-70% charging in 5mins.

BYD recently ran a production Yangwang U7 electric sedan for 30,000km in less than 9 days. It went through more than 350 high-power charging sessions in 9 days, and BYD reported 98.7% battery capacity remaining.

One major fast-charging problem is lithium plating. During charging, lithium ions move through the electrolyte and insert into the graphite anode. Push them in faster than the anode can accept them and some lithium can instead deposit as metallic lithium on its surface. Over time, that consumes usable lithium and reduces capacity.

BYD appears to be attacking that transport bottleneck directly. In Blade Battery 2.0, it says graphite particles in the anode are aligned more perpendicular to the electrode plane. This gives lithium ions a shorter path into the graphite, allowing faster intercalation with lower resistance. BYD also claims improved electrolyte transport and a thinner SEI layer, the protective film through which lithium ions must pass.

Thermal management is equally important. High current generates heat, but a battery that is too cold can make lithium plating worse because ion transport slows down. You need a controlled temperature window, as charge current changes with temperature and state of charge. That I think BYD does well. Fast charging is not determined by charger power or chemistry alone. Electrode geometry, internal resistance, temperature control and the charging profile all matter.

But please note that this result does not directly transfer to your phone or laptop. The same electrochemistry applies, but those cells are built and managed differently. BYD is showing how much faster charging can become when the battery itself is designed around it. They are definitely taking battery tech to new heights.

AdvancedTech #Battery

In an older post we looked at how SynthID hides a watermark inside AI-generated images. Today, let us look at text watermarking and how it works. This week Anthropic says Claude models will use a version of Google DeepMind’s SynthID-Text for all text it generates.

So an LLM writes one token at a time. At each step it usually has several reasonable choices. For eg. “The regulator stayed stable even when the load suddenly _” Here changed, increased or jumped can all fit the line.

SynthID-Text uses this freedom to leave a statistical signature. A secret key, together with the recent text, generates hidden scores for possible next tokens. The sampler chooses from sensible options, and over many steps those choices become correlated with the secret pattern. These scores change with context, so there is no fixed list of “watermark words”. The same token can support the watermark in one sentence and not in another.

A detector with the matching key can later recreate the expected scores and check whether the chosen tokens match that pattern more often than chance. One sentence gives little evidence. Longer passages give more choices to test, so confidence improves. This can be added to many LLMs at the sampling stage without retraining. Each provider can use its own key and configuration. Claude’s detector would therefore look for Claude’s pattern, while another provider may use a different key or method.

The method works best when the model has freedom in wording. Creative writing offers many choices. Short factual answers and code don’t have that many options. Heavy rewriting can also weaken the signal, which is why watermark stealing(deducing patterns from text) is an active research area.

If widely adopted, text watermarking could make AI involvement easier to verify later. I am assuming this will be heavily used in academia and school work. A watermark can indicate that a compatible model likely contributed to the text, without identifying the person or chat behind it.